Feature: speech recognition for visual feedback on audio

[robinp]

Could wire up speech recognition on the audio chunks to:

• auto-name the clips when importing
• show the text an the audio block in the timeline

Bonus: add keyword / topic extraction.

[owickstrom]

Yes, this is something I've been wanting to add. But I haven't found any good tools to integration with yet (maybe haven't looked hard enough...)

[robinp]

You mean speech recognition? I had good experience with CMU PocketSphinx, even wrote some c2hs bindings. If you are interested, I can dig them up!

[owickstrom]

Sorry for dropping the ball on this one. PocketSphinx seems interesting, and it's available on macOS, Linux, and Windows, it seems. Do you have the bindings published somewhere?

[robinp]

Just pushed after some dusting: https://github.com/TreeTide/voicetrans/tree/master/sphinx . Partial bindings, but good enough to run the recognition (see app/Main.hs to get model file and test input).