This is for the assistant reading out the responses. There are various models such as Qwen3, Kokoro, and more.
- Models: https://huggingface.co/models?pipeline_tag=text-to-speech&so...
- Benchmark Blog (with link to the benchmark): https://huggingface.co/blog/open-tts-leaderboard
# ASR = Automated Speech Recognition
This is for the assistant processing speech commands, etc. There are various models for this as well. You will likely get errors in the detected text depending on things like accent and technical/obscure/domain-specific words/terms.
- Models: https://huggingface.co/models?pipeline_tag=automatic-speech-...
- Benchmark: https://huggingface.co/spaces/hf-audio/open_asr_leaderboard
you may find this helpful. I meant speech to text btw
there are dedicated text to speech, speech to text, etc models. check out the space and what's available on huggingface.