How to create audio profiles for speakers for re-identification?

[hmehdi515]

❓ Questions and Help

Any way to create voice profiles?

What is your question?

Is there a way to store and re-identify voice profiles (similar to BOT-Sort for classifying and re-identifying images) by extracting embeddings that characterize the speaker's voice?
Example:
Give model an audio file with containing multiple voices and use speech diarization for segmentation.
Store the embeddings in a structure database for future comparison and re-identification.
Compare new embeddings against stored ones to identify if a speaker has been previously encountered.

What's your environment?

• OS (e.g., Linux): Windows
• FunASR Version (e.g., 1.0.0): 1.0.5