ERROR:root:We're not at the end, yet we recieved less data than expected. THIS SHOULD NOT HAPPEN.

[deama]

I'm having the issue with:

ERROR:root:We're not at the end, yet we recieved less data than expected. THIS SHOULD NOT HAPPEN.

popping up after I try and stream it. Could this be fixed? I'm thinking it needs maybe more time in the buffer before it starts playing it?

Here's my code:

from elevenlabslib import *
import sys
import time

# Initialize the ElevenLabs API
api_key = "[api-key]"
user = ElevenLabsUser(api_key)
voice = user.get_voices_by_name("Adam")[0]

# Audio generation options
playbackOptions = PlaybackOptions(runInBackground=True)
generationOptions = GenerationOptions(model_id="eleven_english_v2", stability=0.5, similarity_boost=1.0, style=0.2, use_speaker_boost=True)

# Generate and play the audio
try:
    audioStreamFuture = voice.generate_stream_audio_v2(sys.argv[1], playbackOptions, generationOptions)[1]
    audioStream = audioStreamFuture.result()

    # Check if the audio stream is ready and active
    while not audioStream.active:
        time.sleep(0.1)  # Wait for the audio stream to become active

    # This loop ensures the audio plays to completion
    while audioStream.active:
        time.sleep(0.1)

except Exception as e:
    print(f"An error occurred: {e}")

[lugia19]

The error suggests you're on an old version of the library - be sure to update to the latest one.

Additionally, here's how you can wait on the audio to finish playing in a better way:

import threading
from elevenlabslib import *
import sys

# Initialize the ElevenLabs API
api_key = "[api-key]"
user = ElevenLabsUser(api_key)
voice = user.get_voices_by_name("Adam")[0]

# Audio generation options
playback_end_event = threading.Event()
playbackOptions = PlaybackOptions(runInBackground=True, onPlaybackEnd=playback_end_event.set)
generationOptions = GenerationOptions(model_id="eleven_english_v2", stability=0.5, similarity_boost=1.0, style=0.2, use_speaker_boost=True)

# Generate and play the audio
try:
    voice.generate_stream_audio_v2(sys.argv[1], playbackOptions, generationOptions)

    #Wait for the audio to play to completion:
    playback_end_event.wait()

except Exception as e:
    print(f"An error occurred: {e}")

[deama]

Ah thanks, that seemed to do the trick, I had to force update it with

pip install --upgrade elevenlabslib

[deama]

Hey sorry to reopen, but now the issue seems to be that it's not closing properly after playing the audio, the python script in task manager remains open until I force close it, here's the full code so far:

import threading
from elevenlabslib import *
import sys

# Initialize the ElevenLabs API
api_key = "[API]"
user = ElevenLabsUser(api_key)
voice = user.get_voices_by_name("Adam")[0]

# Audio generation options
playback_end_event = threading.Event()
playbackOptions = PlaybackOptions(runInBackground=True, onPlaybackEnd=playback_end_event.set)
generationOptions = GenerationOptions(model_id="eleven_english_v2", stability=0.5, similarity_boost=1.0, style=0.2, use_speaker_boost=True)

# Generate and play the audio
try:
    voice.generate_stream_audio_v2(sys.argv[1], playbackOptions, generationOptions)

    #Wait for the audio to play to completion:
    playback_end_event.wait()

except Exception as e:
    print(f"An error occurred: {e}")

Possibly it's due to the library update? It was working fine on the old version 0.10 except for that bug I mentioned earlier that I had before, but on 0.14 it doesn't close?

I don't think it's a threading issue per sey, I tried my old method as well and it gave the same issue.

EDIT: After playing around with it, it appears to get stuck after giving it text that is longer than a specific length, e.g. the following text keeps getting it stuck:

I'm happy for the Iron Tower and all, congratulations on another release. But as a player, to me it's an extremely verbose game. I'll give it another try. But I'll likely drop it. And then wait for the game to become moddable + for ChatGPT to be able to shorten text precisely as instructed, then mod it and play as it should be played, with 90% of the text thrown out.

When it gets stuck, I can also see the memory keep increasing indefinitely.

[lugia19]

That's... certainly interesting. Let me take a look at what's going on.

[lugia19]

I actually can't reproduce this at all - it stops running correctly. Can you set logging level to debug and send over the log?

[deama]

How do set the logging level to debug?

[lugia19]

Add the line logging.basicConfig(level=logging.DEBUG) to the start of the script (you'll need to import logging)

[lugia19]

Okay, nevermind, I have managed to reproduce it - it just took a much longer text. I'll look into it tomorrow and post any updates.

[deama]

It gave this monstrosity:

output.zip

DEBUG:root:Putting 8192 bytes in queue.
DEBUG:root:The following is some logging for a new and fun soundfile bug, which I (poorly) worked around. Lovely.
DEBUG:root:Before the bug: frame 1140480 (byte 413779)
DEBUG:root:After the bug: frame 1140480 (byte 413779)
DEBUG:root:Fun bug happened AFTER the download was over. Continue.
DEBUG:root:Done recreating, now at 1140480 (byte 412525).
DEBUG:root:Read bytes: 8192

What the hell is that?

[lugia19]

Should be fixed in 0.14.2 - give it a shot and let me know.

[deama]

Nice, works, thanks!