You can check out PicoVoice Eagle (paid product): https://picovoice.ai/docs/eagle/ https://picovoice.ai/docs/eagle/ You pass N number of PCM frames through the

You can check out PicoVoice Eagle (paid product): https://picovoice.ai/docs/eagle/ https://picovoice.ai/docs/eagle/ You pass N number of PCM frames through their trainer and once you reach a certain percentage you can extract an embedding you can save. Then you can identify audio against the set of identified speakers and it will return percentage matches for each.