I d need more information like the code

[jade basin]

Yeah ofcourse let me share that

This is the conversion func :

def convert_to_wav(mp3_data):
# Read MP3 audio data into an AudioSegment object
mp3_audio = AudioSegment.from_file(io.BytesIO(mp3_data))
# Create a BytesIO object to hold the WAV data
wav_io = io.BytesIO()
# Export the audio to WAV format into the BytesIO object
mp3_audio.export(wav_io, format="wav")
# Get the WAV byte data
wav_data = wav_io.getvalue()
return wav_data

for audio in elevenlabs.generate(text_gen):
yield convert_to_wav(audio)

[thorn citrus]

so text_gen() is a string iterator, right? So you're doing input streaming?

[jade basin]

yes

[thorn citrus]

and you mentioned you were already decoding the base64 data

I think the problem is that you're assuming each chunk of audio you get back is a complete mp3 file

which it isn't

doing on the fly conversion like that is a lot harder, I believe

my suggestions are:

1. if you specifically need wav audio, use the pcm output_format
2. if you just want to stream the audio, look into some other method

it depends on what you're trying to achieve

[jade basin]

I see so just to summarize even though we are receiving audio chunks, it might be that the chunks are not complete but are partial files.

[thorn citrus]

correct

for example, only the very first chunk will contain the header

(it might be split across multiple chunks even but still, you get the picture)

[jade basin]

Yes now I get the whole picture

btw when you said looking into other methods? what else is there for streaming?

I mean I have come across the generate_stream and generate_stream_input functions

Thank you so much @thorn citrus

[thorn citrus]

oh I didn't mean in the official library

you'd either need to do something custom, or maybe look into how the unofficial elevenlabslib wrapper (mine) does it

which is...complicated, to say the least, but it does work well

like I said, depends on your usecase