*coming soon: use Flow to facilitate self-learning from online videos
python conversation.py to chat with Jojo - he thinks he's funny...
Speechmatics are known for their high-quality multilingual Automatic Speech Recognition. Flow is their conversational API (early access at the time of writing!).
Flow supports speaker diarization, which means that based on acoustic features, it can detect multiple speakers in a conversation and keep track of speaker-specific information!
And you can play around with the persona, style, and context parameters to create your own character!
Turn on/off the transcripts by setting SEE_TRANSCRIPTS = False.
--
Some downsides I noticed:
- LLM component related mistakes. Hard to pinpoint without knowing their architecture exactly, but point is: it misunderstands often.
- Not possible to control voice style during the conversation.
- The default voice amelia is painful to my ears.
--
😮
Check out Flow: https://www.speechmatics.com/flow
API Docs: https://docs.speechmatics.com/flow/getting-started