What is the difference between the encoders for Qwen2 audio, Qwen2.5 Omni, Audio Flamingo 3 and Kimi audio?
#1
by
mifanbushipeicai
- opened
As far as I know, are they all initialized based on Whisper large v3?
In kimi-Audio, whisper encoder is trained. Not freeze. And they use 13 million hours of audio data covering a wide range of modalities including speech, sound, and music.
