TaSTT.git, branch v0.13.2Free self-hosted STT for VRChat.https://git.yummers.dev/TaSTT.git/atom?h=v0.13.22023-07-25T00:52:19+00:00Unity assets can be generated at a configurable path2023-07-25T00:52:19+00:00yumyum.food.vr@gmail.com2023-07-25T00:52:19+00:00urn:sha1:2074749d4263991d835b19257bf1510dcaf55211
Useful for projects with multiple avatars with different animators.
Bugfix: unity panel now shows saved paths2023-07-25T00:27:21+00:00yumyum.food.vr@gmail.com2023-07-25T00:27:21+00:00urn:sha1:e64c71199c39fb1637bb2fd410cf27b9f4575b12
The paths you enter in the Unity panel (animator, menu, params, and
assets folder) are saved in the app config, but were not populated
correctly on app restart or pane redraw. Now they are.
Preserve audio chunk length when dropping samples2023-07-09T02:06:45+00:00yumyum.food.vr@gmail.com2023-07-09T01:55:41+00:00urn:sha1:a602bfb95665697b15a2de58694c6ac064af2916
When we commit a transcription, we drop the corresponding audio data.
Audio data is represented as a list of chunks. Each chunk contains a few
hundred samples of audio data, representing O(10ms) of audio.
If we want to drop a few seconds of data, this means simply deleting
many chunks of audio. There's usually a chunk where we want to drop some
portion of audio data.
Instead of slicing away that part of the chunk, which would change its
length, this change zeroes it out. This preserves the assumption that
each chunk has the same temporal length.
Commit logic now drops parts of frames2023-07-08T22:57:39+00:00yumyum.food.vr@gmail.com2023-07-08T22:57:39+00:00urn:sha1:80f46a7a346e73c94a3bb8ae01099743020ef2a4
We used to drop entire frames only, leading to situations where more
audio is dropped than desired. Now we drop frames down to the precision
of the individual audio sample requested.
Update README2023-07-08T00:54:40+00:00yumyum.food.vr@gmail.com2023-07-08T00:54:40+00:00urn:sha1:5db7426bb14b7e51275c14d8173bd67e8addc4ce
Mostly updating roadmap stuff. Non-VRC use cases are "complete" since I
was mostly targeting streaming. The ability to type into arbitrary text
fields is still somewhat nascent & could be improved.
Also update some other random stuff to be more up to date. KillFrenzy
Avatar Text is now MIT, pog!
Enforce a stricter avg_logbprob than default2023-07-07T09:35:51+00:00yumyum.food.vr@gmail.com2023-07-07T09:30:18+00:00urn:sha1:7a576bcac1c37c3c5a59fadf172aa70b15ff83c8
Common hallucinations sneak in around -0.9 avg_logprob.
Also:
* Limit temperatures to just 0.0. Multiple values cause latency to
occasionally spike.
Filter out segments based on avg_log_prob & no_speech_prob2023-07-07T08:58:45+00:00yumyum.food.vr@gmail.com2023-07-07T08:57:56+00:00urn:sha1:2793ac9dd31059f2fc29f7978bcb688a7de664ed
Surprisingly, these args do not cause transcribe() to omit those
segments from the result, so we have to manually filter them out.
Hallucinated phrases generally have one or both of these params set
high.
Use 16-bit ints with generated silence2023-07-07T08:44:28+00:00yumyum.food.vr@gmail.com2023-07-07T08:44:28+00:00urn:sha1:742eb86d652d7689bbf3ae8b286bf0a6b1c2380d
Each sample of audio data is a 16-bit int, not an 8-bit int.
Fix performance regression2023-07-07T08:27:02+00:00yumyum.food.vr@gmail.com2023-07-07T08:27:02+00:00urn:sha1:cdc4889cb5e752d00f7f8933a5486f4f3441f6e9
Each chunk of audio samples should be encoded as a binary string, not as
a list.
Enforce minimum 5.0 second duration on audio buffer2023-07-07T00:36:14+00:00yumyum.food.vr@gmail.com2023-07-07T00:36:14+00:00urn:sha1:d0d3b18ad0a859e5e7a1cc5b8a569349b505c924
New commit logic would reduce buffer to a size smaller than this,
causing it to hallucinate things like:
* "See you next time!"
* "Thanks for watching!"
* "Bye!"
The hope is that by keeping the buffer at least 5.0 seconds long, as
described in the paper, this will cut down on these events.