Stabilizing Timestamps for Whisper
This library modifies Whisper to produce more reliable timestamps and extends its functionality.
https://github.com/jianfch/stable-ts/assets/28970749/7adf0540-3620-4b2b-b2d4-e316906d6dfa
Setup
Prerequisites: FFmpeg & PyTorch
FFmpeg
Requires [FFmpeg](https://ffmpeg.org/) in PATH ``` # on Ubuntu or Debian sudo apt update && sudo apt install ffmpeg # on Arch Linux sudo pacman -S ffmpeg # on MacOS using Homebrew (https://brew.sh/) brew install ffmpeg # on Windows using Chocolatey (https://chocolatey.org/) choco install ffmpeg # on Windows using Scoop (https://scoop.sh/) scoop install ffmpeg ```PyTorch
If PyTorch is not installed when installing Stable-ts, the default version will be installed which may not have GPU support. To avoid this issue, install your preferred version with instructions at https://pytorch.org/get-started/locally/.pip install -U stable-tsTo install the latest commit:
pip install -U git+https://github.com/jianfch/stable-ts.gitWhisperless Version
To install Stable-ts without Whisper as a dependency: ``` pip install -U stable-ts-whisperless ``` To install the latest Whisperless commit: ``` pip install -U git+https://github.com/jianfch/stable-ts.git@whisperless ```Usage
Transcribe
import stable_whisper
model = stable_whisper.load_model('base')
result = model.transcribe('audio.mp3')
result.to_srt_vtt('audio.srt')CLI
```commandline stable-ts audio.mp3 -o audio.srt ```Docstrings:
load_model()
Load an instance if :class:`whisper.model.Whisper`. Parameters ---------- name : {'tiny', 'tiny.en', 'base', 'base.en', 'small', 'small.en', 'medium', 'medium.en', 'large-v1', 'large-v2', 'large-v3', or 'large'} One of the official model names listed by :func:`whisper.available_models`, or path to a model checkpoint containing the model dimensions and the model state_dict. device : str or torch.device, optional PyTorch device to put the model into. download_root : str, optional Path to download the model files; by default, it uses "~/.cache/whisper". in_memory : bool, default False Whether to preload the model weights into host memory. cpu_preload : bool, default True Load model into CPU memory first then move model to specified device to reduce GPU memory usage when loading model dq : bool, default False Whether to apply Dynamic Quantization to model to reduced memory usage and increase inference speed but at the cost of a slight decrease in accuracy. Only for CPU. engine : str, optional Engine for Dynamic Quantization. Returns ------- model : "Whisper" The Whisper ASR model instance. Notes ----- The overhead from ``dq = True`` might make inference slower for models smaller than 'large'.transcribe()
Transcribe audio using Whisper. This is a modified version of :func:`whisper.transcribe.transcribe` with slightly different decoding logic while allowing additional preprocessing and postprocessing. The preprocessing performed on the audio includes: voice isolation / noise removal and low/high-pass filter. The postprocessing performed on the transcription result includes: adjusting timestamps with VAD and custom regrouping segments based punctuation and speech gaps. Parameters ---------- model : whisper.model.Whisper An instance of Whisper ASR model. audio : str or numpy.ndarray or torch.Tensor or bytes or AudioLoader Path/URL to the audio file, the audio waveform, or bytes of audio file or instance of :class:`stable_whisper.audio.AudioLoader`. If audio is :class:`numpy.ndarray` or :class:`torch.Tensor`, the audio must be already at sampled to 16kHz. verbose : bool or None, default False Whether to display the text being decoded to the console. Displays all the details if ``True``. Displays progressbar if ``False``. Display nothing if ``None``. temperature : float or iterable of float, default (0.0, 0.2, 0.4, 0.6, 0.8, 1.0) Temperature for sampling. It can be a tuple of temperatures, which will be successfully used upon failures according to either ``compression_ratio_threshold`` or ``logprob_threshold``. compression_ratio_threshold : float, default 2.4 If the gzip compression ratio is above this value, treat as failed. logprob_threshold : float, default -1 If the average log probability over sampled tokens is below this value, treat as failed no_speech_threshold : float, default 0.6 If the no_speech probability is higher than this value AND the average log probability over sampled tokens is below ``logprob_threshold``, consider the segment as silent condition_on_previous_text : bool, default True If ``True``, the previous output of the model is provided as a prompt for the next window; disabling may make the text inconsistent across windows, but the model becomes less prone to getting stuck in a failure loop, such as repetition looping or timestamps going out of sync. initial_prompt : str, optional Text to provide as a prompt for the first window. This can be used to provide, or "prompt-engineer" a context for transcription, e.g. custom vocabularies or proper nouns to make it more likely to predict those word correctly. word_timestamps : bool, default True Extract word-level timestamps using the cross-attention pattern and dynamic time warping, and include the timestamps for each word in each segment. Disabling this will prevent segments from splitting/merging properly. regroup : bool or str, default True, meaning the default regroup algorithm String for customizing the regrouping algorithm. False disables regrouping. Ignored if ``word_timestamps = False``. suppress_silence : bool, default True Whether to enable timestamps adjustments based on the detected silence. suppress_word_ts : bool, default True Whether to adjust word timestamps based on the detected silence. Only enabled if ``suppress_silence = True``. use_word_position : bool, default True Whether to use position of the word in its segment to determine whether to keep end or start timestamps if adjustments are required. If it is the first word, keep end. Else if it is the last word, keep the start. q_levels : int, default 20 Quantization levels for generating timestamp suppression mask; ignored if ``vad = true``. Acts as a threshold to marking sound as silent. Fewer levels will increase the threshold of volume at which to mark a sound as silent. k_size : int, default 5 Kernel size for avg-pooling waveform to generate timestamp suppression mask; ignored if ``vad = true``. Recommend 5 or 3; higher sizes will reduce detection of silence. denoiser : str, optional String of the denoiser to use for preprocessing ``audio``. See ``stable_whisper.audio.SUPPORTED_DENOISERS`` for supported denoisers. denoiser_options : dict, optional Options to use for ``denoiser``. vad : bool or dict, default False Whether to use Silero VAD to generate timestamp suppression mask. Instead of ``True``, using a dict of keyword arguments will load the VAD with the arguments. Silero VAD requires PyTorch 1.12.0+. Official repo, https://github.com/snakers4/silero-vad. vad_threshold : float, default 0.35 Threshold for detecting speech with Silero VAD. Low threshold reduces false positives for silence detection. min_word_dur : float or None, default None meaning use ``stable_whisper.default.DEFAULT_VALUES`` Shortest duration each word is allowed to reach for silence suppression. min_silence_dur : float, optional Shortest duration of silence allowed for silence suppression. nonspeech_error : float, default 0.1 Relative error of non-speech sections that appear in between a word for silence suppression. only_voice_freq : bool, default False Whether to only use sound between 200 - 5000 Hz, where majority of human speech are. prepend_punctuations : str or None, default None meaning use ``stable_whisper.default.DEFAULT_VALUES`` Punctuations to prepend to next word. append_punctuations : str or None, default None meaning use ``stable_whisper.default.DEFAULT_VALUES`` Punctuations to append to previous word. stream : bool or None, default None Whether to loading ``audio`` in chunks of 30 seconds until the end of file/stream. If ``None`` and ``audio`` is a string then set to ``True`` else ``False``. mel_first : bool, optional Process entire audio track into log-Mel spectrogram first instead in chunks. Used if odd behavior seen in stable-ts but not in whisper, but use significantly more memory for long audio. split_callback : Callable, optional Custom callback for grouping tokens up with their corresponding words. The callback must take two arguments, list of tokens and tokenizer. The callback returns a tuple with a list of words and a corresponding nested list of tokens. suppress_ts_tokens : bool, default False Whether to suppress timestamp tokens during inference for timestamps are detected at silent. Reduces hallucinations in some cases, but also prone to ignore disfluencies and repetitions. This option is ignored if ``suppress_silence = False``. gap_padding : str, default ' ...' Padding prepend to each segments for word timing alignment. Used to reduce the probability of model predicting timestamps earlier than the first utterance. only_ffmpeg : bool, default False Whether to use only FFmpeg (instead of not yt-dlp) for URls max_instant_words : float, default 0.5 If percentage of instantaneous words in a segment exceed this amount, the segment is removed. avg_prob_threshold: float or None, default None Transcribe the gap after the previous word and if the average word proababiliy of a segment falls below this value, discard the segment. If ``None``, skip transcribing the gap to reduce chance of timestamps starting before the next utterance. nonspeech_skip : float or None, default None Skip non-speech sections that are equal or longer than this duration in seconds. Disable skipping if ``None``. Reduce text and timing hallucinations in non-speech sections but may increase processing time. progress_callback : Callable, optional A function that will be called when transcription progress is updated. The callback need two parameters. The first parameter is a float for seconds of the audio that has been transcribed. The second parameter is a float for total duration of audio in seconds. ignore_compatibility : bool, default False Whether to ignore warnings for compatibility issues with the detected Whisper version. extra_models : list of whisper.model.Whisper, optional List of additional Whisper model instances to use for computing word-timestamps along with ``model``. dynamic_heads : bool or int or str, optional Whether to find optimal cross-attention heads during runtime instead of using the predefined heads for word-timestamp extraction. Specify the number of heads or `True` for default of 6 heads. To specify number of iterations for finding the optimal heads, use string with "," to separate heads and iterations (e.g. "8,3" for 8 heads and 3 iterations). decode_options Keyword arguments to construct class:`whisper.decode.DecodingOptions` instances. Returns ------- stable_whisper.result.WhisperResult All timestamps, words, probabilities, and other data from the transcription of ``audio``. See Also -------- stable_whisper.non_whisper.transcribe_any : Return :class:`stable_whisper.result.WhisperResult` containing all the data from transcribing audio with unmodified :func:`whisper.transcribe.transcribe` with preprocessing and postprocessing. stable_whisper.whisper_word_level.faster_whisper.faster_transcribe : Return :class:`stable_whisper.result.WhisperResult` containing all the data from transcribing audio with :meth:`faster_whisper.WhisperModel.transcribe` with preprocessing and postprocessing. Examples -------- >>> import stable_whisper >>> model = stable_whisper.load_model('base') >>> result = model.transcribe('audio.mp3', vad=True) >>> result.to_srt_vtt('audio.srt') Saved: audio.srttranscribe_minimal()
Transcribe audio using Whisper. This is uses the original whisper transcribe function, :func:`whisper.transcribe.transcribe`, while still allowing additional preprocessing and postprocessing. The preprocessing performed on the audio includes: voice isolation / noise removal and low/high-pass filter. The postprocessing performed on the transcription result includes: adjusting timestamps with VAD and custom regrouping segments based punctuation and speech gaps. Parameters ---------- model : whisper.model.Whisper An instance of Whisper ASR model. audio : str or numpy.ndarray or torch.Tensor or bytes Path/URL to the audio file, the audio waveform, or bytes of audio file. If audio is ``numpy.ndarray`` or ``torch.Tensor``, the audio must be already at sampled to 16kHz. verbose : bool or None, default False Whether to display the text being decoded to the console. Displays all the details if ``True``. Displays progressbar if ``False``. Display nothing if ``None``. word_timestamps : bool, default True Extract word-level timestamps using the cross-attention pattern and dynamic time warping, and include the timestamps for each word in each segment. Disabling this will prevent segments from splitting/merging properly. regroup : bool or str, default True, meaning the default regroup algorithm String for customizing the regrouping algorithm. False disables regrouping. Ignored if ``word_timestamps = False``. suppress_silence : bool, default True Whether to enable timestamps adjustments based on the detected silence. suppress_word_ts : bool, default True Whether to adjust word timestamps based on the detected silence. Only enabled if ``suppress_silence = True``. use_word_position : bool, default True Whether to use position of the word in its segment to determine whether to keep end or start timestamps if adjustments are required. If it is the first word, keep end. Else if it is the last word, keep the start. q_levels : int, default 20 Quantization levels for generating timestamp suppression mask; ignored if ``vad = true``. Acts as a threshold to marking sound as silent. Fewer levels will increase the threshold of volume at which to mark a sound as silent. k_size : int, default 5 Kernel size for avg-pooling waveform to generate timestamp suppression mask; ignored if ``vad = true``. Recommend 5 or 3; higher sizes will reduce detection of silence. denoiser : str, optional String of the denoiser to use for preprocessing ``audio``. See ``stable_whisper.audio.SUPPORTED_DENOISERS`` for supported denoisers. denoiser_options : dict, optional Options to use for ``denoiser``. vad : bool or dict, default False Whether to use Silero VAD to generate timestamp suppression mask. Instead of ``True``, using a dict of keyword arguments will load the VAD with the arguments. Silero VAD requires PyTorch 1.12.0+. Official repo, https://github.com/snakers4/silero-vad. vad_threshold : float, default 0.35 Threshold for detecting speech with Silero VAD. Low threshold reduces false positives for silence detection. min_word_dur : float, default 0.1 Shortest duration each word is allowed to reach for silence suppression. min_silence_dur : float, optional Shortest duration of silence allowed for silence suppression. nonspeech_error : float, default 0.1 Relative error of non-speech sections that appear in between a word for silence suppression. only_voice_freq : bool, default False Whether to only use sound between 200 - 5000 Hz, where majority of human speech are. only_ffmpeg : bool, default False Whether to use only FFmpeg (instead of not yt-dlp) for URls options Additional options used for :func:`whisper.transcribe.transcribe` and :func:`stable_whisper.non_whisper.transcribe_any`. Returns ------- stable_whisper.result.WhisperResult All timestamps, words, probabilities, and other data from the transcription of ``audio``. Examples -------- >>> import stable_whisper >>> model = stable_whisper.load_model('base') >>> result = model.transcribe_minimal('audio.mp3', vad=True) >>> result.to_srt_vtt('audio.srt') Saved: audio.srtfaster-whisper
Use with [faster-whisper](https://github.com/guillaumekln/faster-whisper): ``` pip install -U stable-ts[fw] ``` ```python model = stable_whisper.load_faster_whisper('base') result = model.transcribe_stable('audio.mp3') ``` ```commandline stable-ts audio.mp3 -o audio.srt -fw ``` Docstring:load_faster_whisper()
Load an instance of :class:`faster_whisper.WhisperModel`. Parameters ---------- model_size_or_path : {'tiny', 'tiny.en', 'base', 'base.en', 'small', 'small.en', 'medium', 'medium.en', 'large-v1', 'large-v2', 'large-v3', or 'large'} Size of the model. model_init_options Additional options to use for initialization of :class:`faster_whisper.WhisperModel`. Returns ------- faster_whisper.WhisperModel A modified instance with :func:`stable_whisper.whisper_word_level.load_faster_whisper.faster_transcribe` assigned to :meth:`faster_whisper.WhisperModel.transcribe_stable`.transcribe_stable()
Transcribe audio using faster-whisper (https://github.com/guillaumekln/faster-whisper). This is uses the transcribe method from faster-whisper, :meth:`faster_whisper.WhisperModel.transcribe`, while still allowing additional preprocessing and postprocessing. The preprocessing performed on the audio includes: voice isolation / noise removal and low/high-pass filter. The postprocessing performed on the transcription result includes: adjusting timestamps with VAD and custom regrouping segments based punctuation and speech gaps. Parameters ---------- model : faster_whisper.WhisperModel The faster-whisper ASR model instance. audio : str or numpy.ndarray or torch.Tensor or bytes Path/URL to the audio file, the audio waveform, or bytes of audio file. If audio is :class:`numpy.ndarray` or :class:`torch.Tensor`, the audio must be already at sampled to 16kHz. verbose : bool or None, default False Whether to display the text being decoded to the console. Displays all the details if ``True``. Displays progressbar if ``False``. Display nothing if ``None``. word_timestamps : bool, default True Extract word-level timestamps using the cross-attention pattern and dynamic time warping, and include the timestamps for each word in each segment. Disabling this will prevent segments from splitting/merging properly. regroup : bool or str, default True, meaning the default regroup algorithm String for customizing the regrouping algorithm. False disables regrouping. Ignored if ``word_timestamps = False``. suppress_silence : bool, default True Whether to enable timestamps adjustments based on the detected silence. suppress_word_ts : bool, default True Whether to adjust word timestamps based on the detected silence. Only enabled if ``suppress_silence = True``. use_word_position : bool, default True Whether to use position of the word in its segment to determine whether to keep end or start timestamps if adjustments are required. If it is the first word, keep end. Else if it is the last word, keep the start. q_levels : int, default 20 Quantization levels for generating timestamp suppression mask; ignored if ``vad = true``. Acts as a threshold to marking sound as silent. Fewer levels will increase the threshold of volume at which to mark a sound as silent. k_size : int, default 5 Kernel size for avg-pooling waveform to generate timestamp suppression mask; ignored if ``vad = true``. Recommend 5 or 3; higher sizes will reduce detection of silence. denoiser : str, optional String of the denoiser to use for preprocessing ``audio``. See ``stable_whisper.audio.SUPPORTED_DENOISERS`` for supported denoisers. denoiser_options : dict, optional Options to use for ``denoiser``. vad : bool or dict, default False Whether to use Silero VAD to generate timestamp suppression mask. Instead of ``True``, using a dict of keyword arguments will load the VAD with the arguments. Silero VAD requires PyTorch 1.12.0+. Official repo, https://github.com/snakers4/silero-vad. vad_threshold : float, default 0.35 Threshold for detecting speech with Silero VAD. Low threshold reduces false positives for silence detection. min_word_dur : float or None, default None meaning use ``stable_whisper.default.DEFAULT_VALUES`` Shortest duration each word is allowed to reach for silence suppression. min_silence_dur : float, optional Shortest duration of silence allowed for silence suppression. nonspeech_error : float, default 0.3 Relative error of non-speech sections that appear in between a word for silence suppression. only_voice_freq : bool, default False Whether to only use sound between 200 - 5000 Hz, where majority of human speech are. only_ffmpeg : bool, default False Whether to use only FFmpeg (instead of not yt-dlp) for URls check_sorted : bool, default True Whether to raise an error when timestamps returned by faster-whipser are not in ascending order. progress_callback : Callable, optional A function that will be called when transcription progress is updated. The callback need two parameters. The first parameter is a float for seconds of the audio that has been transcribed. The second parameter is a float for total duration of audio in seconds. options Additional options used for :meth:`faster_whisper.WhisperModel.transcribe` and :func:`stable_whisper.non_whisper.transcribe_any`. Returns ------- stable_whisper.result.WhisperResult All timestamps, words, probabilities, and other data from the transcription of ``audio``. Examples -------- >>> import stable_whisper >>> model = stable_whisper.load_faster_whisper('base') >>> result = model.transcribe_stable('audio.mp3', vad=True) >>> result.to_srt_vtt('audio.srt') Saved: audio.srtHugging Face Transformers (~9x faster)
Run Whisper up to 9x faster with [Hugging Face Transformer](https://huggingface.co/openai/whisper-large-v3): ``` pip install -U stable-ts[hf] ``` ```python model = stable_whisper.load_hf_whisper('base') result = model.transcribe('audio.mp3') ```CLI
```commandline stable-ts audio.mp3 -o audio.srt -hw ```Output
https://github.com/jianfch/stable-ts/assets/28970749/c22dcdf9-79cb-485a-ae38-184d006e513e
Stable-ts supports various text output formats.
result.to_srt_vtt('audio.srt') #SRT
result.to_srt_vtt('audio.vtt') #VTT
result.to_ass('audio.ass') #ASS
result.to_tsv('audio.tsv') #TSVDocstrings:
result_to_srt_vtt()
Generate SRT/VTT from ``result`` to display segment-level and/or word-level timestamp. Parameters ---------- result : dict or list or stable_whisper.result.WhisperResult Result of transcription. filepath : str, default None, meaning content will be returned as a ``str`` Path to save file. segment_level : bool, default True Whether to use segment-level timestamps in output. word_level : bool, default True Whether to use word-level timestamps in output. min_dur : float, default 0.2 Minimum duration allowed for any word/segment before the word/segments are merged with adjacent word/segments. tag: tuple of (str, str), default None, meaning ('', '') if SRT else ('', '') Tag used to change the properties a word at its timestamp. vtt : bool, default None, meaning determined by extension of ``filepath`` or ``False`` if no valid extension. Whether to output VTT. strip : bool, default True Whether to remove spaces before and after text on each segment for output. reverse_text: bool or tuple, default False Whether to reverse the order of words for each segment or provide the ``prepend_punctuations`` and ``append_punctuations`` as tuple pair instead of ``True`` which is for the default punctuations. Returns ------- str String of the content if ``filepath`` is ``None``. Notes ----- ``reverse_text`` will not fix RTL text not displaying tags properly which is an issue with some video player. VLC seems to not suffer from this issue. Examples -------- >>> import stable_whisper >>> model = stable_whisper.load_model('base') >>> result = model.transcribe('audio.mp3') >>> result.to_srt_vtt('audio.srt') Saved: audio.srtresult_to_ass()
Generate Advanced SubStation Alpha (ASS) file from ``result`` to display segment-level and/or word-level timestamp. Parameters ---------- result : dict or list or stable_whisper.result.WhisperResult Result of transcription. filepath : str, default None, meaning content will be returned as a ``str`` Path to save file. segment_level : bool, default True Whether to use segment-level timestamps in output. word_level : bool, default True Whether to use word-level timestamps in output. min_dur : float, default 0.2 Minimum duration allowed for any word/segment before the word/segments are merged with adjacent word/segments. tag: tuple of (str, str) or int, default None, meaning use default highlighting Tag used to change the properties a word at its timestamp. -1 for individual word highlight tag. font : str, default `Arial` Word font. font_size : int, default 48 Word font size. strip : bool, default True Whether to remove spaces before and after text on each segment for output. highlight_color : str, default '00ff00' Hexadecimal of the color use for default highlights as 'result_to_tsv()
Generate TSV from ``result`` to display segment-level and/or word-level timestamp. Parameters ---------- result : dict or list or stable_whisper.result.WhisperResult Result of transcription. filepath : str, default None, meaning content will be returned as a ``str`` Path to save file. segment_level : bool, default True Whether to use segment-level timestamps in output. word_level : bool, default True Whether to use word-level timestamps in output. min_dur : float, default 0.2 Minimum duration allowed for any word/segment before the word/segments are merged with adjacent word/segments. strip : bool, default True Whether to remove spaces before and after text on each segment for output. reverse_text: bool or tuple, default False Whether to reverse the order of words for each segment or provide the ``prepend_punctuations`` and ``append_punctuations`` as tuple pair instead of ``True`` which is for the default punctuations. Returns ------- str String of the content if ``filepath`` is ``None``. Notes ----- ``reverse_text`` will not fix RTL text not displaying tags properly which is an issue with some video player. VLC seems to not suffer from this issue. Examples -------- >>> import stable_whisper >>> model = stable_whisper.load_model('base') >>> result = model.transcribe('audio.mp3') >>> result.to_tsv('audio.tsv') Saved: audio.tsvresult_to_txt()
Generate plain-text without timestamps from ``result``. Parameters ---------- result : dict or list or stable_whisper.result.WhisperResult Result of transcription. filepath : str, default None, meaning content will be returned as a ``str`` Path to save file. min_dur : float, default 0.2 Minimum duration allowed for any word/segment before the word/segments are merged with adjacent word/segments. strip : bool, default True Whether to remove spaces before and after text on each segment for output. reverse_text: bool or tuple, default False Whether to reverse the order of words for each segment or provide the ``prepend_punctuations`` and ``append_punctuations`` as tuple pair instead of ``True`` which is for the default punctuations. Returns ------- str String of the content if ``filepath`` is ``None``. Notes ----- ``reverse_text`` will not fix RTL text not displaying tags properly which is an issue with some video player. VLC seems to not suffer from this issue. Examples -------- >>> import stable_whisper >>> model = stable_whisper.load_model('base') >>> result = model.transcribe('audio.mp3') >>> result.to_txt('audio.txt') Saved: audio.txtsave_as_json()
Save ``result`` as JSON file to ``path``. Parameters ---------- result : dict or list or stable_whisper.result.WhisperResult Result of transcription. path : str Path to save file. ensure_ascii : bool, default False Whether to escape non-ASCII characters. Examples -------- >>> import stable_whisper >>> model = stable_whisper.load_model('base') >>> result = model.transcribe('audio.mp3') >>> result.save_as_json('audio.json') Saved: audio.json
There are word-level and segment-level timestamps. All output formats support them.
They also support will both levels simultaneously except TSV.
By default, segment_level and word_level are both True for all the formats that support both simultaneously.
Examples in VTT.
Default: segment_level=True + word_level=True
CLI
`--segment_level true` + `--word_level true`00:00:07.760 --> 00:00:09.900
But<00:00:07.860> when<00:00:08.040> you<00:00:08.280> arrived<00:00:08.580> at<00:00:08.800> that<00:00:09.000> distant<00:00:09.400> world,segment_level=True + word_level=False
00:00:07.760 --> 00:00:09.900
But when you arrived at that distant world,segment_level=False + word_level=True
00:00:07.760 --> 00:00:07.860
But
00:00:07.860 --> 00:00:08.040
when
00:00:08.040 --> 00:00:08.280
you
00:00:08.280 --> 00:00:08.580
arrived
...JSON
The result can also be saved as a JSON file to preserve all the data for future reprocessing. This is useful for testing different sets of postprocessing arguments without the need to redo inference.
result.save_as_json('audio.json')CLI
```commandline stable-ts audio.mp3 -o audio.json ```Processing JSON file of the results into SRT.
result = stable_whisper.WhisperResult('audio.json')
result.to_srt_vtt('audio.srt')CLI
```commandline stable-ts audio.json -o audio.srt ```Alignment
Audio can be aligned/synced with plain text on word-level.
text = 'Machines thinking, breeding. You were to bear us a new, promised land.'
result = model.align('audio.mp3', text, language='en')When the text is correct but the timestamps need more work,
align() is a faster alternative for testing various settings/models.
new_result = model.align('audio.mp3', result, language='en')CLI
```commandline stable-ts audio.mp3 --align text.txt --language en ``` `--align` can also a JSON file of a resultDocstring:
align()
Align plain text or tokens with audio at word-level. Since this is significantly faster than transcribing, it is a more efficient method for testing various settings without re-transcribing. This is also useful for timing a more correct transcript than one that Whisper can produce. Parameters ---------- model : "Whisper" The Whisper ASR model modified instance audio : str or numpy.ndarray or torch.Tensor or bytes or AudioLoader Path/URL to the audio file, the audio waveform, or bytes of audio file or instance of :class:`stable_whisper.audio.AudioLoader`. If audio is :class:`numpy.ndarray` or :class:`torch.Tensor`, the audio must be already at sampled to 16kHz. text : str or list of int or stable_whisper.result.WhisperResult String of plain-text, list of tokens, or instance of :class:`stable_whisper.result.WhisperResult`. language : str, default None, uses ``language`` in ``text`` if it is a :class:`stable_whisper.result.WhisperResult` Language of ``text``. Required if ``text`` does not contain ``language``. remove_instant_words : bool, default False Whether to truncate any words with zero duration. token_step : int, default 100 Max number of tokens to align each pass. Use higher values to reduce chance of misalignment. original_split : bool, default False Whether to preserve the original segment groupings. Segments are split by line breaks if ``text`` is plain-text. max_word_dur : float or None, default 3.0 Global maximum word duration in seconds. Re-align words that exceed the global maximum word duration. word_dur_factor : float or None, default 2.0 Factor to compute the Local maximum word duration, which is ``word_dur_factor`` * local medium word duration. Words that need re-alignment, are re-algined with duration <= local/global maximum word duration. nonspeech_skip : float or None, default 5.0 Skip non-speech sections that are equal or longer than this duration in seconds. Disable skipping if ``None``. fast_mode : bool, default False Whether to speed up alignment by re-alignment with local/global maximum word duration. ``True`` tends produce better timestamps when ``text`` is accurate and there are no large speechless gaps. tokenizer : "Tokenizer", default None, meaning a new tokenizer is created according ``language`` and ``model`` A tokenizer to used tokenizer text and detokenize tokens. stream : bool or None, default None Whether to loading ``audio`` in chunks of 30 seconds until the end of file/stream. If ``None`` and ``audio`` is a string then set to ``True`` else ``False``. failure_threshold : float, optional Abort alignment when percentage of words with zero duration exceeds ``failure_threshold``. verbose : bool or None, default False Whether to display the text being decoded to the console. Displays all the details if ``True``. Displays progressbar if ``False``. Display nothing if ``None``. regroup : bool or str, default True, meaning the default regroup algorithm String for customizing the regrouping algorithm. False disables regrouping. Ignored if ``word_timestamps = False``. suppress_silence : bool, default True Whether to enable timestamps adjustments based on the detected silence. suppress_word_ts : bool, default True Whether to adjust word timestamps based on the detected silence. Only enabled if ``suppress_silence = True``. use_word_position : bool, default True Whether to use position of the word in its segment to determine whether to keep end or start timestamps if adjustments are required. If it is the first word, keep end. Else if it is the last word, keep the start. q_levels : int, default 20 Quantization levels for generating timestamp suppression mask; ignored if ``vad = true``. Acts as a threshold to marking sound as silent. Fewer levels will increase the threshold of volume at which to mark a sound as silent. k_size : int, default 5 Kernel size for avg-pooling waveform to generate timestamp suppression mask; ignored if ``vad = true``. Recommend 5 or 3; higher sizes will reduce detection of silence. denoiser : str, optional String of the denoiser to use for preprocessing ``audio``. See ``stable_whisper.audio.SUPPORTED_DENOISERS`` for supported denoisers. denoiser_options : dict, optional Options to use for ``denoiser``. vad : bool or dict, default False Whether to use Silero VAD to generate timestamp suppression mask. Instead of ``True``, using a dict of keyword arguments will load the VAD with the arguments. Silero VAD requires PyTorch 1.12.0+. Official repo, https://github.com/snakers4/silero-vad. vad_threshold : float, default 0.35 Threshold for detecting speech with Silero VAD. Low threshold reduces false positives for silence detection. min_word_dur : float or None, default None meaning use ``stable_whisper.default.DEFAULT_VALUES`` Shortest duration each word is allowed to reach for silence suppression. min_silence_dur : float, optional Shortest duration of silence allowed for silence suppression. nonspeech_error : float, default 0.1 Relative error of non-speech sections that appear in between a word for silence suppression. only_voice_freq : bool, default False Whether to only use sound between 200 - 5000 Hz, where majority of human speech are. prepend_punctuations : str or None, default None meaning use ``stable_whisper.default.DEFAULT_VALUES`` Punctuations to prepend to next word. append_punctuations : str or None, default None meaning use ``stable_whisper.default.DEFAULT_VALUES`` Punctuations to append to previous word. progress_callback : Callable, optional A function that will be called when transcription progress is updated. The callback need two parameters. The first parameter is a float for seconds of the audio that has been transcribed. The second parameter is a float for total duration of audio in seconds. ignore_compatibility : bool, default False Whether to ignore warnings for compatibility issues with the detected Whisper version. extra_models : list of whisper.model.Whisper, optional List of additional Whisper model instances to use for computing word-timestamps along with ``model``. presplit : bool or list of str, default True meaning ['.', '。', '?', '？'] List of ending punctuation used to split ``text`` into segments for applying ``gap_padding``, but segmentation of final output is unnaffected unless ``original_split=True``. If ``original_split=True``, the original split is used instead of split from ``presplit``. Ignored if ``model`` is a faster-whisper model. gap_padding : str, default ' ...' Only if ``presplit=True``, ``gap_padding`` is prepended to each segments for word timing alignment. Used to reduce the probability of model predicting timestamps earlier than the first utterance. Ignored if ``model`` is a faster-whisper model. dynamic_heads : bool or int or str, optional Whether to find optimal cross-attention heads during runtime instead of using the predefined heads for word-timestamp extraction. Specify the number of heads or `True` for default of 6 heads. To specify number of iterations for finding the optimal heads, use string with "," to separate heads and iterations (e.g. "8,3" for 8 heads and 3 iterations). Returns ------- stable_whisper.result.WhisperResult or None All timestamps, words, probabilities, and other data from the alignment of ``audio``. Return None if alignment fails and ``remove_instant_words = True``. Notes ----- If ``token_step`` is less than 1, ``token_step`` will be set to its maximum value, 442. This value is computed with ``whisper.model.Whisper.dims.n_text_ctx`` - 6. IF ``original_split = True`` and a line break is found in middle of a word in ``text``, the split will occur after that word. ``regroup`` is ignored if ``original_split = True``. Examples -------- >>> import stable_whisper >>> model = stable_whisper.load_model('base') >>> result = model.align('helloworld.mp3', 'Hello, World!', 'English') >>> result.to_srt_vtt('helloword.srt') Saved 'helloworld.srt'Adjustments
Timestamps are adjusted after the model predicts them.
When suppress_silence=True (default), transcribe()/transcribe_minimal()/align() adjust based on silence/non-speech.
The timestamps can be further adjusted base on another result with adjust_by_result(),
which acts as a logical AND operation for the timestamps of both results, further reducing duration of each word.
Note: both results are required to have word timestamps and matching words.
# the adjustments are in-place for `result`
result.adjust_by_result(new_result)Docstring:
adjust_by_result()
Minimize the duration of words using timestamps of another result. Parameters ---------- other_result : "WhisperResult" Timing data of the same words in a WhisperResult instance. min_word_dur : float or None, default None meaning use ``stable_whisper.default.DEFAULT_VALUES`` Prevent changes to timestamps if the resultant word duration is less than ``min_word_dur``. verbose : bool, default False Whether to print out the timestamp changes.Refinement
Timestamps can be further improved with refine().
This method iteratively mutes portions of the audio based on current timestamps
then compute the probabilities of the tokens.
Then by monitoring the fluctuation of the probabilities, it tries to find the most precise timestamps.
"Most precise" in this case means the latest start and earliest end for the word
such that it still meets the specified conditions.
model.refine('audio.mp3', result)CLI
```commandline stable-ts audio.mp3 --refine -o audio.srt ``` Input can also be JSON file of a result. ```commandline stable-ts result.json --refine -o audio.srt --refine_option "audio=audio.mp3" ```Docstring:
refine()
Improve existing timestamps. This function iteratively muting portions of the audio and monitoring token probabilities to find the most precise timestamps. This "most precise" in this case means the latest start and earliest end of a word that maintains an acceptable probability determined by the specified arguments. This is useful readjusting timestamps when they start too early or end too late. Parameters ---------- model : "Whisper" The Whisper ASR model modified instance audio : str or numpy.ndarray or torch.Tensor or bytes Path/URL to the audio file, the audio waveform, or bytes of audio file. If audio is :class:`numpy.ndarray` or :class:`torch.Tensor`, the audio must be already at sampled to 16kHz. result : stable_whisper.result.WhisperResult All timestamps, words, probabilities, and other data from the transcription of ``audio``. steps : str, default 'se' Instructions for refinement. A 's' means refine start-timestamps. An 'e' means refine end-timestamps. rel_prob_decrease : float, default 0.3 Maximum percent decrease in probability relative to original probability which is the probability from muting according initial timestamps. abs_prob_decrease : float, default 0.05 Maximum decrease in probability from original probability. rel_rel_prob_decrease : float, optional Maximum percent decrease in probability relative to previous probability which is the probability from previous iteration of muting. prob_threshold : float, default 0.5 Stop refining the timestamp if the probability of its token goes below this value. rel_dur_change : float, default 0.5 Maximum percent change in duration of a word relative to its original duration. abs_dur_change : float, optional Maximum seconds a word is allowed deviate from its original duration. word_level : bool, default True Whether to refine timestamps on word-level. If ``False``, only refine start/end timestamps of each segment. precision : float, default 0.1 Precision of refined timestamps in seconds. The lowest precision is 0.02 second. single_batch : bool, default False Whether to process in only batch size of one to reduce memory usage. inplace : bool, default True Whether to alter timestamps in-place. Return a deepcopy of ``result`` if ``False``. denoiser : str, optional String of the denoiser to use for preprocessing ``audio``. See ``stable_whisper.audio.SUPPORTED_DENOISERS`` for supported denoisers. denoiser_options : dict, optional Options to use for ``denoiser``. only_voice_freq : bool, default False Whether to only use sound between 200 - 5000 Hz, where majority of human speech are. verbose : bool or None, default False Whether to display the text being decoded to the console. Displays all the details if ``True``. Displays progressbar if ``False``. Display nothing if ``None``. Returns ------- stable_whisper.result.WhisperResult All timestamps, words, probabilities, and other data from the refinement of ``text`` with ``audio``. Notes ----- The lower the ``precision``, the longer the processing time. Examples -------- >>> import stable_whisper >>> model = stable_whisper.load_model('base') >>> result = model.transcribe('audio.mp3') >>> model.refine('audio.mp3', result) >>> result.to_srt_vtt('audio.srt') Saved 'audio.srt'Regrouping Words
Stable-ts has a preset for regrouping words into different segments with more natural boundaries.
This preset is enabled by regroup=True (default).
But there are other built-in regrouping methods that allow you to customize the regrouping algorithm.
This preset is just a predefined combination of those methods.
https://github.com/jianfch/stable-ts/assets/28970749/7b6164a3-50e2-4368-8b75-853cb14045ec
# The following results are all functionally equivalent:
result0 = model.transcribe('audio.mp3', regroup=True) # regroup is True by default
result1 = model.transcribe('audio.mp3', regroup=False)
(
result1
.clamp_max()
.split_by_punctuation([(',', ' '), '，'])
.split_by_gap(.5)
.merge_by_gap(.3, max_words=3)
.split_by_punctuation([('.', ' '), '。', '?', '？'])
)
result2 = model.transcribe('audio.mp3', regroup='cm_sp=,* /，_sg=.5_mg=.3+3_sp=.* /。/?/？')
# To undo all regrouping operations:
result0.reset()Any regrouping algorithm can be expressed as a string. Please feel free share your strings here
Regrouping Methods
regroup()
Regroup (in-place) words into segments. Parameters ---------- regroup_algo: str or bool, default 'da' String representation of a custom regrouping algorithm or ``True`` use to the default algorithm 'da'. verbose : bool, default False Whether to show all the methods and arguments parsed from ``regroup_algo``. only_show : bool, default False Whether to show the all methods and arguments parsed from ``regroup_algo`` without running the methods Returns ------- stable_whisper.result.WhisperResult The current instance after the changes. Notes ----- Syntax for string representation of custom regrouping algorithm. Method keys: sg: split_by_gap sp: split_by_punctuation sl: split_by_length sd: split_by_duration mg: merge_by_gap mp: merge_by_punctuation ms: merge_all_segment cm: clamp_max l: lock us: unlock_all_segments da: default algorithm (cm_sp=,* /，_sg=.5_mg=.3+3_sp=.* /。/?/？) rw: remove_word rs: remove_segment rp: remove_repetition rws: remove_words_by_str fg: fill_in_gaps p: pad ag: adjust_gaps csl: convert_to_segment_leve Metacharacters: = separates a method key and its arguments (not used if no argument) _ separates method keys (after arguments if there are any) + separates arguments for a method key / separates an argument into list of strings * separates an item in list of strings into a nested list of strings Notes: -arguments are parsed positionally -if no argument is provided, the default ones will be used -use 1 or 0 to represent True or False Example 1: merge_by_gap(.2, 10, lock=True) mg=.2+10+++1 Note: [lock] is the 5th argument hence the 2 missing arguments inbetween the three + before 1 Example 2: split_by_punctuation([('.', ' '), '。', '?', '？'], True) sp=.* /。/?/？+1 Example 3: merge_all_segments().split_by_gap(.5).merge_by_gap(.15, 3) ms_sg=.5_mg=.15+3split_by_gap()
Split (in-place) any segment where the gap between two of its words is greater than ``max_gap``. Parameters ---------- max_gap : float, default 0.1 Maximum second(s) allowed between two words if the same segment. lock : bool, default False Whether to prevent future splits/merges from altering changes made by this method. newline: bool, default False Whether to insert line break at the split points instead of splitting into separate segments. Returns ------- stable_whisper.result.WhisperResult The current instance after the changes.split_by_punctuation()
Split (in-place) segments at words that start/end with ``punctuation``. Parameters ---------- punctuation : list of str or list of tuple of (str, str) or str Punctuation(s) to split segments by. lock : bool, default False Whether to prevent future splits/merges from altering changes made by this method. newline : bool, default False Whether to insert line break at the split points instead of splitting into separate segments. min_words : int, optional Split segments with words >= ``min_words``. min_chars : int, optional Split segments with characters >= ``min_chars``. min_dur : int, optional split segments with duration (in seconds) >= ``min_dur``. Returns ------- stable_whisper.result.WhisperResult The current instance after the changes.split_by_length()
Split (in-place) any segment that exceeds ``max_chars`` or ``max_words`` into smaller segments. Parameters ---------- max_chars : int, optional Maximum number of characters allowed in each segment. max_words : int, optional Maximum number of words allowed in each segment. even_split : bool, default True Whether to evenly split a segment in length if it exceeds ``max_chars`` or ``max_words``. force_len : bool, default False Whether to force a constant length for each segment except the last segment. This will ignore all previous non-locked segment boundaries. lock : bool, default False Whether to prevent future splits/merges from altering changes made by this method. include_lock: bool, default False Whether to include previous lock before splitting based on max_words, if ``even_split = False``. Splitting will be done after the first non-locked word > ``max_chars`` / ``max_words``. newline: bool, default False Whether to insert line break at the split points instead of splitting into separate segments. Returns ------- stable_whisper.result.WhisperResult The current instance after the changes. Notes ----- If ``even_split = True``, segments can still exceed ``max_chars`` and locked words will be ignored to avoid uneven splitting.split_by_duration()
Split (in-place) any segment that exceeds ``max_dur`` into smaller segments. Parameters ---------- max_dur : float Maximum duration (in seconds) per segment. even_split : bool, default True Whether to evenly split a segment in length if it exceeds ``max_dur``. force_len : bool, default False Whether to force a constant length for each segment except the last segment. This will ignore all previous non-locked segment boundaries. lock : bool, default False Whether to prevent future splits/merges from altering changes made by this method. include_lock: bool, default False Whether to include previous lock before splitting based on max_words, if ``even_split = False``. Splitting will be done after the first non-locked word > ``max_dur``. newline: bool, default False Whether to insert line break at the split points instead of splitting into separate segments. Returns ------- stable_whisper.result.WhisperResult The current instance after the changes. Notes ----- If ``even_split = True``, segments can still exceed ``max_dur`` and locked words will be ignored to avoid uneven splitting.merge_by_gap()
Merge (in-place) any pair of adjacent segments if the gap between them <= ``min_gap``. Parameters ---------- min_gap : float, default 0.1 Minimum second(s) allow between two segment. max_words : int, optional Maximum number of words allowed in each segment. max_chars : int, optional Maximum number of characters allowed in each segment. is_sum_max : bool, default False Whether ``max_words`` and ``max_chars`` is applied to the merged segment instead of the individual segments to be merged. lock : bool, default False Whether to prevent future splits/merges from altering changes made by this method. newline : bool, default False Whether to insert a line break between the merged segments. Returns ------- stable_whisper.result.WhisperResult The current instance after the changes.merge_by_punctuation()
Merge (in-place) any two segments that has specific punctuations inbetween. Parameters ---------- punctuation : list of str or list of tuple of (str, str) or str Punctuation(s) to merge segments by. max_words : int, optional Maximum number of words allowed in each segment. max_chars : int, optional Maximum number of characters allowed in each segment. is_sum_max : bool, default False Whether ``max_words`` and ``max_chars`` is applied to the merged segment instead of the individual segments to be merged. lock : bool, default False Whether to prevent future splits/merges from altering changes made by this method. newline : bool, default False Whether to insert a line break between the merged segments. Returns ------- stable_whisper.result.WhisperResult The current instance after the changes.merge_all_segments()
Merge all segments into one segment. Returns ------- stable_whisper.result.WhisperResult The current instance after the changes.clamp_max()
Clamp all word durations above certain value. This is most effective when applied before and after other regroup operations. Parameters ---------- medium_factor : float, default 2.5 Clamp durations above (``medium_factor`` * medium duration) per segment. If ``medium_factor = None/0`` or segment has less than 3 words, it will be ignored and use only ``max_dur``. max_dur : float, optional Clamp durations above ``max_dur``. clip_start : bool or None, default None Whether to clamp the start of a word. If ``None``, clamp the start of first word and end of last word per segment. verbose : bool, default False Whether to print out the timestamp changes. Returns ------- stable_whisper.result.WhisperResult The current instance after the changes.lock()
Lock words/segments with matching prefix/suffix to prevent splitting/merging. Parameters ---------- startswith: str or list of str Prefixes to lock. endswith: str or list of str Suffixes to lock. right : bool, default True Whether prevent splits/merges with the next word/segment. left : bool, default False Whether prevent splits/merges with the previous word/segment. case_sensitive : bool, default False Whether to match the case of the prefixes/suffixes with the words/segments. strip : bool, default True Whether to ignore spaces before and after both words/segments and prefixes/suffixes. Returns ------- stable_whisper.result.WhisperResult The current instance after the changes.pad()
Pad (in-place) timestamps in chronological order. Parameters ---------- start_pad : float, optional Seconds to pad start timestamps. Each start timestamp will be extended no earlier than the end timestamp of the previous word. end_pad : float, optional Seconds to pad end timestamps. Each end timestamp will be extended no later than the start timestamp of the next word or ``max_end``. max_dur : float, optional Only pad segments or words (``word_level=True``) with duration (in seconds) under or equal to ``max_dur``. max_end : float, optional Timestamp (in seconds) that padded timestamps cannot exceed. Generally used to prevent the last padded end timestamp from exceeding the total duration of the audio. word_level : bool, default False Whether to pad segment timestamps or word timestamps. Returns ------- stable_whisper.result.WhisperResult The current instance after the changes.Editing
The editing methods in stable-ts can be chained with Regrouping Methods and used in regroup().
Remove specific instances words or segments:
# Remove first word of the first segment:
first_word = result[0][0]
result.remove_word(first_word)
# This following is also does the same:
del result[0][0]
# Remove the last segment:
last_segment = result[-1]
result.remove_segment(last_segment)
# This following is also does the same:
del result[-1]Docstrings:
remove_word()
Remove a word. Parameters ---------- word : WordTiming or tuple of (int, int) Instance of :class:`stable_whisper.result.WordTiming` or tuple of (segment index, word index). reassign_ids : bool, default True Whether to reassign segment and word ids (indices) after removing ``word``. verbose : bool, default True Whether to print detail of the removed word. Returns ------- stable_whisper.result.WhisperResult The current instance after the changes.remove_segment()
Remove a segment. Parameters ---------- segment : Segment or int Instance :class:`stable_whisper.result.Segment` or segment index. reassign_ids : bool, default True Whether to reassign segment IDs (indices) after removing ``segment``. verbose : bool, default True Whether to print detail of the removed word. Returns ------- stable_whisper.result.WhisperResult The current instance after the changes.Removing repetitions:
# Example 1: "This is is is a test." -> "This is a test."
# The following removes the last two " is":
result.remove_repetition(1)
# Example 2: "This is is is a test this is a test." -> "This is a test."
# The following removes the second " is" and third " is", then remove the last "this is a test"
# The first parameter `max_words` is `4` because "this is a test" consists 4 words
result.remove_repetition(4)Docstring:
remove_repetition()
Remove words that repeat consecutively. Parameters ---------- max_words : int Maximum number of words to look for consecutively. case_sensitive : bool, default False Whether the case of words need to match to be considered as repetition. strip : bool, default True Whether to ignore spaces before and after each word. ignore_punctuations : bool, default '"',.?!' Ending punctuations to ignore. extend_duration: bool, default True Whether to extend the duration of the previous word to cover the duration of the repetition. verbose: bool, default True Whether to print detail of the removed repetitions. Returns ------- stable_whisper.result.WhisperResult The current instance after the changes.Removing specific word(s) by string content:
# Remove all " ok" from " ok ok this is a test."
result.remove_words_by_str('ok')
# Remove all " ok" and " Um..." from " ok this is a test. Um..."
result.remove_words_by_str(['ok', 'um'])Docstring:
remove_words_by_str()
Remove words that match ``words``. Parameters ---------- words : str or list of str or None A word or list of words to remove.``None`` for all words to be passed into ``filters``. case_sensitive : bool, default False Whether the case of words need to match to be considered as repetition. strip : bool, default True Whether to ignore spaces before and after each word. ignore_punctuations : bool, default '"',.?!' Ending punctuations to ignore. min_prob : float, optional Acts as the first filter the for the words that match ``words``. Words with probability < ``min_prob`` will be removed if ``filters`` is ``None``, else pass the words into ``filters``. Words without probability will be treated as having probability < ``min_prob``. filters : Callable, optional A function that takes an instance of :class:`stable_whisper.result.WordTiming` as its only argument. This function is custom filter for the words that match ``words`` and were not caught by ``min_prob``. verbose: Whether to print detail of the removed words. Returns ------- stable_whisper.result.WhisperResult The current instance after the changes.Filling in segment gaps:
# result0: [" How are you?"] [" I'm good."] [" Good!"]
# result1: [" Hello!"] [" How are you?"] [" How about you?"] [" Good!"]
result0.fill_in_gaps(result1)
# After filling in the gaps in `result0` with contents in `result1`:
# result0: [" Hello!"] [" How are you?"] [" I'm good."] [" How about you?"] [" Good!"]Docstring:
fill_in_gaps()
Fill in segment gaps larger than ``min_gap`` with content from ``other_result`` at the times of gaps. Parameters ---------- other_result : WhisperResult or str Another transcription result as an instance of :class:`stable_whisper.result.WhisperResult` or path to the JSON of the result. min_gap : float, default 0.1 The minimum seconds of a gap between segments that must be exceeded to be filled in. case_sensitive : bool, default False Whether to consider the case of the first and last word of the gap to determine overlapping words to remove before filling in. strip : bool, default True Whether to ignore spaces before and after the first and last word of the gap to determine overlapping words to remove before filling in. ignore_punctuations : bool, default '"',.?!' Ending punctuations to ignore in the first and last word of the gap to determine overlapping words to remove before filling in. verbose: Whether to print detail of the filled content. Returns ------- stable_whisper.result.WhisperResult The current instance after the changes.Locating Words
There are two ways to locate words. The first way is by approximating time at which the words are spoken then transcribing a few seconds around the approximated time. This also the faster way for locating words.
matches = model.locate('audio.mp3', 'are', language='en', count=0)
for match in matches:
print(match.to_display_str())
# verbose=True does the same thing as this for-loop.Docstring:
locate()
Locate when specific words are spoken in ``audio`` without fully transcribing. This is usefully for quickly finding at what time the specify words or phrases are spoken in an audio. Since it does not need to transcribe the audio to approximate the time, it is significantly faster transcribing then locating the word in the transcript. It can also transcribe few seconds around the approximated time to find out what was said around those words or confirm if the word was even spoken near that time. Parameters ---------- model : whisper.model.Whisper An instance of Whisper ASR model. audio : str or numpy.ndarray or torch.Tensor or bytes Path/URL to the audio file, the audio waveform, or bytes of audio file. If audio is :class:`numpy.ndarray` or :class:`torch.Tensor`, the audio must be already at sampled to 16kHz. text: str or list of int Words/phrase or list of tokens to search for in ``audio``. language : str Language of the ``text``. count : int, default 1, meaning stop search after 1 match Number of matches to find. Use 0 to look for all. duration_window : float or tuple of (float, float), default 3.0, same as (3.0, 3.0) Seconds before and after the end timestamp approximations to transcribe after mode 1. If tuple pair of values, then the 1st value will be seconds before the end and 2nd value will be seconds after. mode : int, default 0 Mode of search. 2, Approximates the end timestamp of ``text`` in the audio. This mode does not confirm whether ``text`` is spoken at the timestamp 1, Completes mode 2 then transcribes audio within ``duration_window`` to confirm whether `text` is a match at the approximated timestamp by checking if ``text`` at that ``duration_window`` is within ``probability_threshold`` or matching the string content if ``text`` with the transcribed text at the ``duration_window``. 0, Completes mode 1 then add word timestamps to the transcriptions of each match. Modes from fastest to slowest: 2, 1, 0 start : float, optional, meaning it starts from 0s Seconds into the audio to start searching for ``text``. end : float, optional Seconds into the audio to stop searching for ``text``. probability_threshold : float, default 0.5 Minimum probability of each token in ``text`` for it to be considered a match. eots : int, default 1 Number of EOTs to reach before stopping transcription at mode 1. When transcription reach a EOT, it usually means the end of the segment or audio. Once ``text`` is found in the ``duration_window``, the transcription will stop immediately upon reaching a EOT. max_token_per_seg : int, default 20 Maximum number of tokens to transcribe in the ``duration_window`` before stopping. exact_token : bool, default False Whether to find a match base on the exact tokens that make up ``text``. case_sensitive : bool, default False Whether to consider the case of ``text`` when matching in string content. verbose : bool or None, default False Whether to display the text being decoded to the console. Displays all the details if ``True``. Displays progressbar if ``False``. Display nothing if ``None``. initial_prompt : str, optional Text to provide as a prompt for the first window. This can be used to provide, or "prompt-engineer" a context for transcription, e.g. custom vocabularies or proper nouns to make it more likely to predict those word correctly. suppress_tokens : str or list of int, default '-1', meaning suppress special characters except common punctuations List of tokens to suppress. denoiser : str, optional String of the denoiser to use for preprocessing ``audio``. See ``stable_whisper.audio.SUPPORTED_DENOISERS`` for supported denoisers. denoiser_options : dict, optional Options to use for ``denoiser``. only_voice_freq : bool, default False Whether to only use sound between 200 - 5000 Hz, where majority of human speech are. Returns ------- stable_whisper.result.Segment or list of dict or list of float Mode 0, list of instances of :class:`stable_whisper.result.Segment`. Mode 1, list of dictionaries with end timestamp approximation of matches and transcribed neighboring words. Mode 2, list of timestamps in seconds for each end timestamp approximation. Notes ----- For ``text``, the case and spacing matters as 'on', ' on', ' On' are different tokens, therefore chose the one that best suits the context (e.g. ' On' to look for it at the beginning of a sentence). Use a sufficiently large first value of ``duration_window`` i.e. the value > time it is expected to speak ``text``. If ``exact_token = False`` and the string content matches, then ``probability_threshold`` is not used. Examples -------- >>> import stable_whisper >>> model = stable_whisper.load_model('base') >>> matches = model.locate('audio.mp3', 'are', language='English', verbose=True) Some words can sound the same but have different spellings to increase of the chance of finding such words use ``initial_prompt``. >>> matches = model.locate('audio.mp3', ' Nickie', 'English', verbose=True, initial_prompt='Nickie')CLI
``` stable-ts audio.mp3 --locate "are" --language en -to "count=0" ```The second way allows you to locate words with regular expression, but it requires the audio to be fully transcribed first.
result = model.transcribe('audio.mp3')
# Find every sentence that contains "and"
matches = result.find(r'[^.]+and[^.]+\.')
# print the all matches if there are any
for match in matches:
print(f'match: {match.text_match}\n'
f'text: {match.text}\n'
f'start: {match.start}\n'
f'end: {match.end}\n')
# Find the word before and after "and" in the matches
matches = matches.find(r'\s\S+\sand\s\S+')
for match in matches:
print(f'match: {match.text_match}\n'
f'text: {match.text}\n'
f'start: {match.start}\n'
f'end: {match.end}\n')Docstring:
find()
Find segments/words and timestamps with regular expression. Parameters ---------- pattern : str RegEx pattern to search for. word_level : bool, default True Whether to search at word-level. flags : optional RegEx flags. Returns ------- stable_whisper.result.WhisperResultMatches An instance of :class:`stable_whisper.result.WhisperResultMatches` with word/segment that match ``pattern``.Silence Suppression
While the timestamps predicted by Whisper are generally accurate,
it sometimes predicts the start of a word way before the word is spoken
or the end of a word long after the word has been spoken.
This is where "silence suppression" helps. It is enabled by default (suppress_silence=True).
The idea is to adjust the timestamps based on the timestamps of non-speech portions of the audio.
Note: In 1.X, "silence suppression" refers to the process of suppressing timestamp tokens of the silent portions during inference,
but changed to post-inference timestamp adjustments in 2.X, which allows stable-ts to be used with other ASR models.
The timestamp token suppression feature is disabled by default, but can still be enabled with suppress_ts_tokens=True.
By default, stable-ts determines the non-speech timestamps based on
how loud a section of the audio is relative to the neighboring sections.
This method is most effective for cases, where the speech is significantly louder than the background noise.
The other method is to use Silero VAD (enabled with vad=True).
To visualize the differences between non-VAD and VAD, see Visualizing Suppression.
Besides the parameters for non-speech detection sensitivity (see Visualizing Suppression),
the following parameters are used to combat inaccurate non-speech detection.
min_word_dur is the shortest duration each word is allowed from adjustments.
nonspeech_error is the relative error of the non-speech that appears in between a word.
use_word_position is whether to use word position in segment to determine whether to keep end or start timestamps
Note: nonspeech_error was not available before 2.14.0; use_word_position was not available before 2.14.2;
min_word_dur prevented any adjustments that resulted in word duration shorter than min_word_dur.
For the following example, min_word_dur=0.5 (default: 0.1) and nonspeech_error=0.3 (default: 0.3).
nonspeech_error=0.3 allows each non-speech section to be treated 1.3 times their actual duration.
Either from the start of the corresponding word to the end of the non-speech
or from the start of the non-speech to the end of the corresponding word.
In the case that both conditions are met, the shorter one is used.
Or if both are equal, then the start of the non-speech to the end of the word is used.
The second non-speech from 1.375s to 1.75s is ignored for 'world.' because it failed both conditions.
The first word, 'Hello', satisfies only the former condition from 0s to 0.625, thus the new start for 'Hello'
would be 0.625s. However, min_word_dur=0.5 requires the resultant duration to be at least 0.5s.
As a result, the start of 'Hello' is changed to 0.375s instead of 0.625s.
Furthermore, the default setting, use_word_position=True, also ensures the start is adjusted for the first word
and the end is adjusted for the last word of the segment as long as one of the conditions is true.
Gap Adjustment
Generally, the initial timestamps are accurate enough for captioning/subtitling purposes. However, starting/ending 200ms earlier/later than the actual time the words are actually spoken makes the timestamps not viable for tasks that require greater accuracy, such as audio editing.
Gap Adjustment makes use of timestamps that accurately if not exactly marks the boundary of utterances.
These timestamps, nonspeech_sections, are computed during transcription / alignment and used for Silence Suppression.
Since the initial timestamps are not pinpoint accuracy, they tend to overlap with nonspeech_sections
which is what the approach exploits.
Unlike Silence Suppression which only clips timestamps,
Gap Adjustment sets the segment timestamps to the timestamps of an ideal section in nonspeech_sections
(i.e. nonspeech start to segment end and nonspeech end to segment start).
This only applies to segment timestamps because there more prominent gaps between segments than between words. Therefore, it is recommended to use Gap Adjustment before any regrouping operations that split segments at places that are unlikely to have gaps. And the opposite is recommended to be applied (e.g. splitting by gaps or punctuations that indicate a pause) before Gap Adjustment .
result = model.transcribe('audio.mp3')
result.adjust_gaps()_Note: .adjust_gaps() can be chained with regrouping methods_
The reliability of the nonspeech_sections varies by audio and settings.
It is generally recommended to use vad=False for audio with no sound in the gaps and vad=True otherwise.
Their respective sensitivities can be tuned. See Visualizing Suppression.
If the result contains word-level data (i.e. word-timestamps),
only nonspeech_sections that overlap the duration of the gap and words adjacent to the gap are considered to avoid clipping
other words of the segments.
There are cases, such as when timestamps are less accurate than usual, where the ideal nonspeech_sections does not overlap the aforementioned duration.
In such cases, it recommended to apply Gap Adjustment after converting the result from word-level to segment-level.
result.convert_to_segment_level().adjust_gaps()Note: .convert_to_segment_level() can be chained with regrouping methods
Docstring:
adjust_gaps()
Set segment timestamps to the timestamps of overlapping``nonspeech_sections`` within ``duration_threshold``. For results with segments split by gaps or punctuations that indicate pauses in speech, the timestamps of these gaps are replaced with those in ``nonspeech_sections`` which overlap. These timestamps tend to be more accurate if not exact timestamps of the start and end of utterances, especially for audio with no sounds in the gaps. Parameters ---------- duration_threshold : float, default 0.75 Only ``nonspeech_sections`` with duration >= ``duration_threshold`` * duration of the longest section that overlaps the existing gap between segments are considered for that gap. one_section : bool, default False Whether to use only one section of ``nonspeech_sections`` for each gap between segments. Recommend ``True`` when there are no sounds in the gaps. Returns ------- stable_whisper.result.WhisperResult The current instance after the changes. Notes ----- The reliability of this approach hinges on the accuracy of the ``nonspeech_sections`` which is computed when ``vad`` is ``False`` or ``True`` for transcription and alignment. The accuracy increases with cleaner audio. Results produced by ``vad=True`` will work better with less clean audio (i.e. audio with non-speech sounds).convert_to_segment_level()
Remove all word-level data. Converting from word-level to segment-level is useful when the word-level data restricts certain operations that need to be performed at the segment-level or when the segment-level results of operations are more desirable. Returns ------- stable_whisper.result.WhisperResult The current instance after the changes. Notes ----- Since many regrouping operations require word-timestamps, it is recommended to only use this method after the result has been regrouped into its final segments.Tips
- do not disable word timestamps with
word_timestamps=Falsefor reliable segment timestamps - use
vad=Truefor more accurate non-speech detection - use
denoiser="demucs"to isolate vocals with Demucs; it is also effective at isolating vocals even if there is no music - use
denoiser="demucs"andvad=Truefor music - set same seed for each transcription (e.g.
random.seed(0)) fordenoiser="demucs"to produce deterministic outputs - to enable dynamic quantization for inference on CPU use
--dq truefor CLI ordq=Trueforstable_whisper.load_model - use
encode_video_comparison()to encode multiple transcripts into one video for synced comparison; see Encode Comparison - use
visualize_suppression()to visualize the differences between non-VAD and VAD options; see Visualizing Suppression - refinement can an effective (but slow) alternative for polishing timestamps if silence suppression isn't effective
- use
--persist/-pfor CLI to keep the CLI running without reloading the same model after it finishes executing its commands
Visualizing Suppression
You can visualize which parts of the audio will likely be suppressed (i.e. marked as silent).
The computed timestamps for any result can be found in result.nonspeech_sections.
Requires: Pillow or opencv-python.
Without VAD
import stable_whisper
# regions on the waveform colored red are where it will likely be suppressed and marked as silent
# ``q_levels=20`` and ``k_size=5`` (default)
stable_whisper.visualize_suppression('audio.mp3', 'image.png', q_levels=20, k_size = 5) 
With Silero VAD
# ``vad_threshold=0.35`` (default)
stable_whisper.visualize_suppression('audio.mp3', 'image.png', vad=True, vad_threshold=0.35)
Docstring:
visualize_suppression()
Visualize regions on the waveform of ``audio`` detected as silent. Regions on the waveform colored red are detected as silent. Parameters ---------- audio : str or numpy.ndarray or torch.Tensor or bytes Path/URL to the audio file, the audio waveform, or bytes of audio file. If audio is ``numpy.ndarray`` or ``torch.Tensor``, the audio must be already at sampled to 16kHz. output : str, default None, meaning image will be shown directly via Pillow or opencv-python Path to save visualization. q_levels : int, default 20 Quantization levels for generating timestamp suppression mask; ignored if ``vad = true``. Acts as a threshold to marking sound as silent. Fewer levels will increase the threshold of volume at which to mark a sound as silent. k_size : int, default 5 Kernel size for avg-pooling waveform to generate timestamp suppression mask; ignored if ``vad = true``. Recommend 5 or 3; higher sizes will reduce detection of silence. vad : bool or dict, default False Whether to use Silero VAD to generate timestamp suppression mask. Instead of ``True``, using a dict of keyword arguments will load the VAD with the arguments. Silero VAD requires PyTorch 1.12.0+. Official repo, https://github.com/snakers4/silero-vad. vad_threshold : float, default 0.35 Threshold for detecting speech with Silero VAD. Low threshold reduces false positives for silence detection. max_width : int, default 1500 Maximum width of visualization to avoid overly large image from long audio. Each unit of pixel is equivalent to 1 token. Use -1 to visualize the entire audio track. height : int, default 200 Height of visualization.Encode Comparison
You can encode videos similar to the ones in the doc for comparing transcriptions of the same audio.
stable_whisper.encode_video_comparison(
'audio.mp3',
['audio_sub1.srt', 'audio_sub2.srt'],
output_videopath='audio.mp4',
labels=['Example 1', 'Example 2']
)Docstring:
encode_video_comparison()
Encode multiple subtitle files into one video with the subtitles vertically stacked. Parameters ---------- audiofile : str Path of audio file. subtitle_files : list of str List of paths for subtitle file. output_videopath : str, optional Output video path. labels : list of str, default, None, meaning use ``subtitle_files`` as labels List of labels for ``subtitle_files``. height : int, default 90 Height for each subtitle section. width : int, default 720 Width for each subtitle section. color : str, default 'black' Background color of the video. fontsize: int, default 70 Font size for subtitles. border_color : str, default 'white' Border color for separating the sections of subtitle. label_color : str, default 'white' Color of labels. label_size : int, default 14 Font size of labels. fps : int, default 25 Frame-rate of the video. video_codec : str, optional Video codec opf the video. audio_codec : str, optional Audio codec opf the video. overwrite : bool, default False Whether to overwrite existing video files with the same path as the output video. only_cmd : bool, default False Whether to skip encoding and only return the full command generate from the specified options. verbose : bool, default True Whether to display ffmpeg processing info. Returns ------- str or None Encoding command as a string if ``only_cmd = True``.Multiple Files with CLI
Transcribe multiple audio files then process the results directly into SRT files.
stable-ts audio1.mp3 audio2.mp3 audio3.mp3 -o audio1.srt audio2.srt audio3.srtAny ASR
You can use most of the features of Stable-ts improve the results of any ASR model/APIs. Just follow this notebook.
Quick 1.X → 2.X Guide
What's new in 2.0.0?
- updated to use Whisper's more reliable word-level timestamps method.
- the more reliable word timestamps allow regrouping all words into segments with more natural boundaries.
- can now suppress silence with Silero VAD (requires PyTorch 1.12.0+)
- non-VAD silence suppression is also more robust
Usage changes
results_to_sentence_srt(result, 'audio.srt')→result.to_srt_vtt('audio.srt', word_level=False)results_to_word_srt(result, 'audio.srt')→result.to_srt_vtt('output.srt', segment_level=False)results_to_sentence_word_ass(result, 'audio.srt')→result.to_ass('output.ass')- there's no need to stabilize segments after inference because they're already stabilized during inference
transcribe()returns aWhisperResultobject which can be converted todictwith.to_dict(). e.gresult.to_dict()
License
This project is licensed under the MIT License - see the LICENSE file for details
Acknowledgments
Includes slight modification of the original work: Whisper