RealtimeSTT

RealtimeSTT copied to clipboard

Is your feature request related to a problem? Please describe. Support for real time streaming STT (Fireworks) for Speech recognition

Describe the solution you'd like Implement Fireworks in Realtime streaming modes via the APIs: https://fireworks.ai/models/fireworks/streaming-speech

Describe alternatives you've considered I intensively looked at paid services like Whisper Flow and especially Aqua Voice (which is awesome!!!) https://withaqua.com/. They are great but expensive with around 10usd per month. An Api based alternative would be a huge added benefit for the witsy community . Aqua voice really made headlines in hacker news just recently ( https://news.ycombinator.com/item?id=39828686 )

Additional context You should try Aqua voice, it has a very nice unintrusive interface and the push and hold to talk shortcut feels natural, even more so than waiting for silence. It also supports custom vocabulary, which is handy and could be implemented via Gladia.

Fireworks seems to have a great price performance ratio. You can try it for free in this playground, do not take my word for it: https://fireworks.ai/models/fireworks/streaming-speech/playground It is just 19 Cents / hour and thus around a third of the cost of the others: https://fireworks.ai/blog/streaming-audio-launch

@KoljaB So, what do you think about using Fireworks Streaming STT?

Playground somehow did not work for me. Also it would require me to isolate the transcription parts. Currently the lib is free, MIT and does not involve external APIs. I don't think I'll put in the work soon needed to extract transcription.