TaSTT.git/Scripts, branch v0.20.0 Free self-hosted STT for VRChat. https://git.yummers.dev/TaSTT.git/atom?h=v0.20.0 2024-07-12T23:10:05+00:00 Another edge case: first commit should not get a leading space 2024-07-12T23:10:05+00:00 yum yum.food.vr@gmail.com 2024-07-12T23:10:05+00:00 urn:sha1:59578a647c9e259b6eb70cd9bee5428fae2bd0b9 Edge case: initial preview should not have a space added in front of it 2024-07-12T23:08:22+00:00 yum yum.food.vr@gmail.com 2024-07-12T23:08:22+00:00 urn:sha1:d7307f8b4d7a8871f08c5999e551b9f58d256c1c God this code is a fucking nightmare Translation shows original language by default 2024-07-12T22:41:23+00:00 yum yum.food.vr@gmail.com 2024-07-12T22:41:23+00:00 urn:sha1:426d0139273a3112ed18f1b7c39eb75b9278fbe1 * Add checkbox to disable this feature if so desired. * Delete old optimization code; can get it back from git if needed. * Enforce that there's at least one space character ' ' between committed segments. Fix translation plugin 2024-07-12T21:48:36+00:00 yum yum.food.vr@gmail.com 2024-07-12T21:48:36+00:00 urn:sha1:75069522ffc8863a356d95e509c81612a3703458 Translation needs torch to convert the nllb model, but the latest version (2.3.1) has an embedded OMP dll which clashes with ctranslate2's dll. Using the last minor version instead (2.2.2) doesn't clash. Also propagate the device, quantization, and flash attention settings to the translator. If you're using GPU, this is a HUUUUGE performance uplift. Translation is basically instant. The bigger models are now feasible to use. Bump CUDNN to v8.9.7 2024-06-09T23:43:34+00:00 yum yum.food.vr@gmail.com 2024-06-09T23:43:34+00:00 urn:sha1:4fec36c3cc00bd649dfb3c9d7e9079b5c8685a0e Also disable flash-attention when CPU mode is selected Add checkbox for flash-attention 2024-06-09T22:54:30+00:00 yum yum.food.vr@gmail.com 2024-06-09T22:54:30+00:00 urn:sha1:f2b21dd5afebd6b76b5835168f7d1bd3bec21f5d Pre-3000 series GPUs don't support it. Oops! Upgrade faster-whisper with flash-attention2 2024-06-06T01:15:47+00:00 yum yum.food.vr@gmail.com 2024-06-06T01:15:47+00:00 urn:sha1:4f0fb5b17de990517e3c1de7ffee5d0f3c9a8961 This should be significantly more efficient than prior versions. * add large-v3 & distilled variant * simplify model acquisition code now that distilled models are part of faster-whisper. Fix distilled models 2024-03-15T01:03:54+00:00 yum yum.food.vr@gmail.com 2024-03-15T01:03:54+00:00 urn:sha1:5638d86c97041de31217e058e411034143e9c882 These were broken due to some logic errors in the codepath which acquires models from huggingface. Distilled large-v2 seems promising as a new default model. Add "simple" text-to-text demo for the modular avatar chatbox 2024-03-09T02:21:22+00:00 yum yum.food.vr@gmail.com 2024-03-09T02:21:22+00:00 urn:sha1:cdc079fb59832fce46708df36ac80ede6d2bd046 To use it: $ python3 -m pip install python-osc pillow $ cd Scripts $ python3 ./text_to_text_demo.py Finish plumbing GPU compute type 2024-02-10T01:51:53+00:00 yum yum.food.vr@gmail.com 2024-02-10T01:51:53+00:00 urn:sha1:5ef207d28f2a9d943384b9ec6872aedae2917ac0