TaSTT.git/Scripts/transcribe_v2.py, branch v0.17.0Free self-hosted STT for VRChat.https://git.yummers.dev/TaSTT.git/atom?h=v0.17.02023-12-09T02:13:56+00:00Add distilled whisper large-v2 model2023-12-09T02:13:56+00:00yumyum.food.vr@gmail.com2023-12-09T02:13:32+00:00urn:sha1:15368618a109eeec69209a6693839eb359ecd190Add distilled whisper-medium model2023-11-07T23:05:29+00:00yumyum.food.vr@gmail.com2023-11-07T23:05:29+00:00urn:sha1:dbb2f72792e2af3ff220313f84bf76a9a1ddbeb4
I converted distil-whisper-medium.en to CTranslate2 format and uploaded
it to huggingface. This model is exceptionally fast and light compared
to the non-distilled version, at the cost of some accuracy.
Transcripts preceding long pauses now drop2023-10-06T01:28:42+00:00yumyum.food.vr@gmail.com2023-10-06T01:22:55+00:00urn:sha1:add7bd8ef86ec21cd1327eb45bcb739aa54f7db8
When hot-miking into the built-in chatbox, there are sometimes long
pauses in conversation. After these pauses, it's undesirable to show the
transcript generate before the pause. This feature makes it so that
those transcripts can be dropped.
Also:
* Limit number of segments sent to browser source to 10. Allow this to
grow up to 10 segments before dropping the first 5 segments.
* Silence warnings generated by `install_in_venv`, used by e.g.
translation codepath.
* Enable audio normalization to improve accuracy when speaking softly,
at the cost of some accuracy when speaking normally.
Credit: user endo0269 on Discord suggested this feature.
Reimplement BrowserSource as a StreamingPlugin2023-09-19T04:23:14+00:00yumyum.food.vr@gmail.com2023-09-19T04:00:56+00:00urn:sha1:c2bc70c18d2fd1c3601b32f2a93b3b4a704786a5
BrowserSource now fades text out continuously over time.
TODO
* Delete C++ webserver, browsersource, transcript code
* Add UI for text age fading
Bugfixes2023-09-16T22:49:55+00:00yumyum.food.vr@gmail.com2023-09-16T22:49:55+00:00urn:sha1:d4c85f4ac4cb627e2611359d18615d76eda29c90
* uwu filter no longer adds extra whitespace before/after segments. This
would defeat commit logic.
* disabling phonemes works again - path to prefab was being quoted
twice, breaking the codepath.
Bugfix: list input devices works again2023-09-12T22:33:18+00:00yumyum.food.vr@gmail.com2023-09-12T22:33:18+00:00urn:sha1:8fcc6c248554a0b08ecd4b43cc0971b78810c080
Oops :)
Pin huggingface_hub to 0.16.42023-09-11T20:53:14+00:00yumyum.food.vr@gmail.com2023-09-11T20:46:57+00:00urn:sha1:0447f37fb744a1b350f6b92e4d140dbdb1c8d3ec
0.17.x are breaking faster_whisper's ability to download models.
Also:
* Start using frozen requirements.txt.
* Conditionally install torch & legacy whisper only when doing
mechanical optimization.
Introduce notion of PresentationFilter2023-09-11T05:52:52+00:00yumyum.food.vr@gmail.com2023-09-11T05:51:16+00:00urn:sha1:d3c325c4c4dd954a75267b013f33f5f3c5d041bc
... and restructure RemoveTrailingPeriod as a filter instead of as a
plugin.
Plugins have the power to change transcription data as it comes along,
but don't have access to the entire transcript. Filters have access to
the entire transcript but can't durably change it.
TODO
* This does not work with data passed through OSC
Fix paging bug2023-09-11T01:33:08+00:00yumyum.food.vr@gmail.com2023-09-11T01:29:59+00:00urn:sha1:920d6dfeeac132488c85311512fe9e5da505c4a8
OSC was paging using incorrect board resolution. Use cfg to provide this
data.
Fix local audio indicators2023-09-10T22:41:25+00:00yumyum.food.vr@gmail.com2023-09-10T22:41:25+00:00urn:sha1:4a4909919223a7446944c6248472c7f71a30307c