SRT output

[ghost]

I seen the other, issue but this definitely needed i dont think json to srt conv will work good as pure as whisper's normal srt option

[SeeknnDestroy]

which issue are you referring to @toprak ? is this model not able to save SRT?

[ghost]

@SeeknnDestroy yeah it only gives json

[SeeknnDestroy]

have you tried using the writer from whisperx library @toprak ? https://github.com/m-bain/whisperX/blob/main/whisperx/utils.py#L406C10-L406C10

[SeeknnDestroy]

Here is how they use their writer https://github.com/m-bain/whisperX/blob/main/whisperx/transcribe.py#L155C8-L155C8

[Vaibhavs10]

Hey @toprak @SeeknnDestroy - That's a lovely suggestion, would either of you mind helping with this? I am currently quite swamped with work and would appreciate any help possible!

[souvikqb]

@toprak This API provides SRT output with Whisper-v3 - https://developer.monsterapi.ai/reference/post_generate-speech2text-v2

The output is same as that given by faster-whisper

[edward-io]

Quick script to convert the json into srt format, courtesy of chatGPT. Tested with VLC.

import json

def convert_to_srt(chunks):
    """
    Convert the chunks of transcripts into .srt format with correct newline placements.
    """
    srt_format = []

    for index, chunk in enumerate(chunks, start=1):
        start_time = chunk['timestamp'][0]
        end_time = chunk['timestamp'][1]

        # Convert timestamp to SRT format: hours:minutes:seconds,milliseconds
        start_srt = f"{int(start_time // 3600):02d}:{int(start_time % 3600 // 60):02d}:{int(start_time % 60):02d},{int(start_time % 1 * 1000):03d}"
        end_srt = f"{int(end_time // 3600):02d}:{int(end_time % 3600 // 60):02d}:{int(end_time % 60):02d},{int(end_time % 1 * 1000):03d}"

        # Append the formatted string to the list, with correct newline placements
        srt_format.append(f"{index}\n{start_srt} --> {end_srt}\n{chunk['text'].strip()}\n\n")

    return srt_format

# Path to the JSON file
file_path = '/path/to/your/transcript.json'

# Load the JSON file
with open(file_path, 'r') as file:
    transcript_data = json.load(file)

# Convert the transcript data to .srt format
srt_data = convert_to_srt(transcript_data['chunks'])

# Path for the output .srt file
srt_file_path = '/path/to/save/converted_transcript.srt'

# Writing to a .srt file
with open(srt_file_path, 'w') as file:
    file.writelines(srt_data)

print(f"Converted .srt file saved at: {srt_file_path}")