Commit graph

  • ca6f9246cc force language = en for .en models Quentin Fuxa 2025-09-28 11:04:00 +0200
  • d080d675a8 cutom alignment heads parameter for custom models Quentin Fuxa 2025-09-27 11:04:00 +0200
  • 40bff38933
    Merge pull request #239 from msghik/feature/fine-tuned-model-support Quentin Fuxa 2025-09-29 10:08:26 +0200
  • 8cbaeecc75 cutom alignment heads parameter for custom models Quentin Fuxa 2025-09-27 11:04:00 +0200
  • 2fe3ca0188 connect source to output destination when used as chrome extension to keep audio playing Quentin Fuxa 2025-09-27 13:59:44 +0200
  • 545ea15c9a ensure buffer size to be a multiple of the element size Quentin Fuxa 2025-09-27 13:58:32 +0200
  • 70e854b346 feat: Allow loading fine-tuned models in simulstreaming google-labs-jules[bot] 2025-09-27 07:29:30 +0000
  • cd9a32a36b update archi to show fastapi server is independent from core Quentin Fuxa 2025-09-21 11:04:00 +0200
  • 6caf3e0485 correct silence handling in translation Quentin Fuxa 2025-09-27 11:58:00 +0200
  • d55490cd27 typo and simpler conditions Quentin Fuxa 2025-09-26 20:38:26 +0200
  • 1fa9e1f656
    Merge pull request #238 from CorentinvdBdO/fix_install Quentin Fuxa 2025-09-26 20:35:29 +0200
  • 994f30e1ed fix: translation in pyproject cvandenbroek 2025-09-26 20:08:35 +0200
  • b22478c0b4 correct silences handling when language not auto Quentin Fuxa 2025-09-25 23:20:00 +0200
  • 94c34efd90 chrome extension ws default to localhost Quentin Fuxa 2025-09-25 23:04:00 +0200
  • 32099b9275 demo extension Quentin Fuxa 2025-09-25 23:59:24 +0200
  • 9fc6654a4a common frontend for web/ and chrome extension Quentin Fuxa 2025-09-25 23:14:25 +0200
  • d24c110d55 to 0.2.11 Quentin Fuxa 2025-09-24 22:34:01 +0200
  • 4dd5d8bf8a translation compatible with auto and detected language Quentin Fuxa 2025-09-22 11:20:00 +0200
  • 93f002cafb language detection after few seconds working Quentin Fuxa 2025-09-20 11:08:00 +0200
  • c5e30c2c07 svg loaded once in javascript, no more need for StaticFiles Quentin Fuxa 2025-09-20 11:06:00 +0200
  • 1c2afb8bd2 svg loaded once in javascript, no more need for StaticFiles Quentin Fuxa 2025-09-20 11:06:00 +0200
  • 674b20d3af in buffer while language not detected » Quentin Fuxa 2025-09-21 11:05:00 +0200
  • a5503308c5 O(n) to O(1) for simulstreaming timestamp determination Quentin Fuxa 2025-09-21 11:04:00 +0200
  • e61afdefa3 punctuation is now checked in timed_object Quentin Fuxa 2025-09-22 22:40:39 +0200
  • 426d70a790 simulstreaming infer does not return a dictionary anymore Quentin Fuxa 2025-09-21 11:03:00 +0200
  • b03a212fbf fixes #227 , auto language dectection v0.1 - simulstreaming only - when diarization and auto Quentin Fuxa 2025-09-19 19:15:28 +0200
  • 1833e7c921 0.2.10 Quentin Fuxa 2025-09-16 23:45:00 +0200
  • 777ec63a71 --pcm-input option information Quentin Fuxa 2025-09-17 16:06:28 +0200
  • 0a6e5ae9c1 ffmpeg install instruction error indicates --pcm-input alternative Quentin Fuxa 2025-09-16 23:32:00 +0200
  • ee448a37e9 when pcm-input is set, the frontend uses AudioWorklet Quentin Fuxa 2025-09-16 23:51:00 +0200
  • 9c051052b0 Merge branch 'main' into ScriptProcessorNode-to-AudioWorklet Quentin Fuxa 2025-09-17 11:28:36 +0200
  • 4d7c487614 replace deprecated ScriptProcessorNode with AudioWorklet Quentin Fuxa 2025-09-16 23:31:00 +0200
  • 65025cc448 nllb backend can be transformers, and model size can be 1.3B Quentin Fuxa 2025-09-16 23:30:00 +0200
  • bbba1d9bb7 add nllb-backend and translation perf test in dev_notes Quentin Fuxa 2025-09-16 20:45:01 +0200
  • 99dc96c644 fixes #224 Quentin Fuxa 2025-09-14 17:03:00 +0200
  • 2a27d2030a feat: support web audio 16kHz PCM input and remove ffmpeg dependency GeorgeCaoJ 2025-09-15 21:29:55 +0800
  • cd160caaa1 asyncio.to_thread for transcription and translation Quentin Fuxa 2025-09-13 22:06:00 +0200
  • d27b5eb23e
    Merge pull request #219 from notV3NOM/main Quentin Fuxa 2025-09-15 10:19:26 +0200
  • f9d704a900 Merge branch 'main' of https://github.com/notv3nom/whisperlivekit into pr/notV3NOM/219 Quentin Fuxa 2025-09-15 10:00:14 +0200
  • 2f6e00f512 simulstreaming warmup is done in whisperlivekit.simul_whisper.backend.load_model, not in warmup_online Quentin Fuxa 2025-09-15 09:43:15 +0200
  • 5aa312e437 simulstreaming warmup is done in whisperlivekit.simul_whisper.backend.load_model, not in warmup_online Quentin Fuxa 2025-09-13 20:19:19 +0100
  • ebaf36a8be
    Fix warmup file behavior notV3NOM 2025-09-13 20:44:24 +0530
  • babe93b99a to 0.2.9 Quentin Fuxa 2025-09-11 21:36:32 +0200
  • a4e9f3cab7 support for raw PCM input option by @YeonjunNotFR Quentin Fuxa 2025-09-11 21:32:11 +0200
  • b06866877a add --disable-punctuation-split option Quentin Fuxa 2025-09-11 21:03:00 +0200
  • 967cdfebc8 fix Translation imports Quentin Fuxa 2025-09-11 21:03:00 +0200
  • 3c11c60126 fix by @treeaaa Quentin Fuxa 2025-09-11 21:03:00 +0200
  • 2963e8a757 translate when at least 3 new tokens Quentin Fuxa 2025-09-09 21:45:00 +0200
  • cb2d4ea88a audio processor lines use now Lines objects instead of dict Quentin Fuxa 2025-09-09 21:45:00 +0200
  • add7ea07ee translator takes all the tokens from the queue Quentin Fuxa 2025-09-09 19:55:39 +0200
  • da8726b2cb
    Merge pull request #211 from Alexander-ARTV/main Quentin Fuxa 2025-09-09 15:46:59 +0200
  • 3358877054 Fix StorageView conversion for CPU/GPU compatibility Quentin Fuxa 2025-09-09 15:44:16 +0200
  • 1f7798c7c1 condition on encoder_feature_ctranslate type Quentin Fuxa 2025-09-09 12:16:52 +0200
  • c7b3bb5e58 Fix regression with faster-whisper encoder_feature Alexander Lindberg 2025-09-09 11:18:55 +0300
  • f661f21675 translation asyncio task Quentin Fuxa 2025-09-07 16:30:00 +0200
  • b6164aa59b translation device determined with torch.device Quentin Fuxa 2025-09-08 11:34:40 +0200
  • 4209d7f7c0 Place all tensors on the same device in sortformer diarization Quentin Fuxa 2025-09-08 10:20:57 +0200
  • 334b338ab0 use platform to determine system and recommand mlx whisper Quentin Fuxa 2025-09-07 15:49:11 +0200
  • 72f33be6f2 translation: use of get_nllb_code Quentin Fuxa 2025-09-07 15:25:14 +0200
  • 84890b8e61
    Merge pull request #201 from notV3NOM/main Quentin Fuxa 2025-09-07 15:18:54 +0200
  • c6668adcf3
    Merge pull request #200 from notV3NOM/misc Quentin Fuxa 2025-09-07 15:17:42 +0200
  • a178ed5c22
    fix simulstreaming preload model count argument in cli notV3NOM 2025-09-06 18:18:09 +0530
  • 7601c74c9c
    add vram usage for large-v3-turbo notV3NOM 2025-09-06 17:56:39 +0530
  • fad9ee4d21
    Merge pull request #198 from notV3NOM/main Quentin Fuxa 2025-09-05 20:46:36 +0200
  • d1a9913c47 nllb v0 Quentin Fuxa 2025-09-04 23:12:00 +0200
  • e4ca2623cb
    Fix scrolling UX with sticky header controls notV3NOM 2025-09-05 21:25:13 +0530
  • 9c1bf37960 fixes #197 Quentin Fuxa 2025-09-05 16:34:13 +0200
  • f46528471b revamp chromium extension settings Quentin Fuxa 2025-09-05 16:19:48 +0200
  • 191680940b Merge branch 'main' of https://github.com/QuentinFuxa/whisper_streaming_web Quentin Fuxa 2025-09-04 23:58:51 +0200
  • ee02afec56 workaround to get the list of microphones in the extension Quentin Fuxa 2025-09-04 23:58:48 +0200
  • a458028de2
    Merge pull request #196 from notV3NOM/main Quentin Fuxa 2025-09-04 23:05:59 +0200
  • abd8f2c269
    Fix exponentially growing simulstreaming silence timer notV3NOM 2025-09-04 21:49:07 +0530
  • f3ad4e39e4 torch.Tensor to torch.as_tensor Quentin Fuxa 2025-09-03 23:01:00 +0200
  • e0a5cbf0e7 v0.1.0 chrome extension Quentin Fuxa 2025-09-03 22:25:00 +0200
  • 953697cd86 torch.Tensor to torch.as_tensor Quentin Fuxa 2025-09-03 20:11:00 +0200
  • 3bd2122eb4 0.2.8 : only the decoder of whisper is loaded in memory when a different encoder is used Quentin Fuxa 2025-09-02 21:12:25 +0200
  • 50b0527858 update architecture Quentin Fuxa 2025-09-01 21:24:12 +0200
  • b044fcdec2 Merge branch 'main' of https://github.com/QuentinFuxa/whisper_streaming_web Quentin Fuxa 2025-09-01 14:55:19 +0200
  • b0508fcf2c mlx/fasterWhisper encoders are loaded once and shared in simulstreaming Quentin Fuxa 2025-08-31 14:55:11 +0200
  • ce89b0aebc
    Merge pull request #177 from komiyamma/translate-readme-to-japanese Quentin Fuxa 2025-09-01 13:54:50 +0200
  • d5008ed828 mlx/fasterWhisper encoders are loaded once and shared in simulstreaming Quentin Fuxa 2025-08-31 12:33:19 +0200
  • d467716e26 add microphone picker Quentin Fuxa 2025-08-31 10:12:52 +0200
  • 199e21b3ef faster-whisper as an optional encoder alternative for simulstreaming Quentin Fuxa 2025-08-30 23:50:16 +0200
  • 1d926f2e67 mlx-whisper used as simulstreaming encoder: improve speed for macos systems Quentin Fuxa 2025-08-30 22:19:11 +0200
  • 4a71a391b8 get_web_interface_html to get_inline_ui_html for embedded web interface HTML Quentin Fuxa 2025-08-30 13:44:06 +0200
  • d3ed4e46e2 Translate README.md to Japanese google-labs-jules[bot] 2025-08-30 04:16:18 +0000
  • 057a1026d7 Merge branch 'main' of https://github.com/QuentinFuxa/whisper_streaming_web Quentin Fuxa 2025-08-29 22:01:04 +0200
  • 1ba171a58d add embedded web interface HTML (single-file version with inline CSS/JS/SVG) Quentin Fuxa 2025-08-29 21:58:51 +0200
  • 1adac67155
    explanations about model persistency in containers Quentin Fuxa 2025-08-29 21:27:08 +0200
  • 42be1a3773
    Merge pull request #173 from CoderRahul9904/chore/docker/pytorch-timeout-retries Quentin Fuxa 2025-08-29 21:22:30 +0200
  • 0a49fafa0d
    Update Dockerfile Rahul Mourya 2025-08-30 00:23:59 +0530
  • 4a5d5e1f3b raise Exception when language == auto and task == translation Quentin Fuxa 2025-08-29 17:44:46 +0200
  • 583a2ec2e4 highlight Sortformer optional installation Quentin Fuxa 2025-08-27 21:02:25 +0200
  • 19765e89e9 remove triton <3 condition Quentin Fuxa 2025-08-27 20:44:39 +0200
  • 9895bc83bf auto detection of language for warmup if not indicated Quentin Fuxa 2025-08-27 20:37:48 +0200
  • ab98c31f16 trim will happen before audio processor Quentin Fuxa 2025-08-27 18:17:11 +0200
  • f9c9c4188a optional dependencies removed, ask to direct alternative package installations Quentin Fuxa 2025-08-27 18:15:32 +0200
  • c21d2302e7 to 0.2.7 Quentin Fuxa 2024-08-24 19:28:00 +0200
  • 4ed62e181d when silences are detected, speaker correction is no more applied Quentin Fuxa 2024-08-24 19:24:00 +0200
  • 52a755a08c indications on how to choose a model Quentin Fuxa 2024-08-24 19:22:00 +0200