Commits

Commit history

1. Commits on Oct 3, 2026

   • Add OWSM v4 and OWSM-CTC v4 (#774)

     Show description for 2ba9fa8
     0xShug0

     authored
     2ba9fa8View commit details

     Copy full SHA for 2ba9fa8

     View code at this pointBrowse repository at this point
2. Commits on Oct 2, 2026

   • Add Audio Flamingo 3 and Next support (#763)

     Show description for a83d302
     0xShug0

     authored
     a83d302View commit details

     Copy full SHA for a83d302

     View code at this pointBrowse repository at this point

   • Add CrisperWhisper 2.0 with long-form transcription and alignment

     0xShug0

     committed
     f9228b6View commit details

     Copy full SHA for f9228b6

     View code at this pointBrowse repository at this point
3. Commits on Oct 1, 2026

   • Add Index-Echo S2TT translation models (#755)

     0xShug0

     authored
     76e26b8View commit details

     Copy full SHA for 76e26b8

     View code at this pointBrowse repository at this point

   • Add KugelAudio native TTS with batched CFG and streaming (#744)

     0xShug0

     authored
     7c2ec18View commit details

     Copy full SHA for 7c2ec18

     View code at this pointBrowse repository at this point

   • Add Smart Turn v3.2 offline turn detection (#743)

     0xShug0

     authored
     8ebcc04View commit details

     Copy full SHA for 8ebcc04

     View code at this pointBrowse repository at this point

   • Add Sidon speech restoration model support (#742)

     0xShug0

     authored
     a83ed6aView commit details

     Copy full SHA for a83ed6a

     View code at this pointBrowse repository at this point
4. Commits on Sep 29, 2026

   • Organize framework components by responsibility (#734)

     0xShug0

     authored
     ed96b73View commit details

     Copy full SHA for ed96b73

     View code at this pointBrowse repository at this point
5. Commits on Sep 28, 2026

   • Make model-local component architecture names explicit (3) (#729)

     Show description for f825d1d
     0xShug0

     authored
     f825d1dView commit details

     Copy full SHA for f825d1d

     View code at this pointBrowse repository at this point

   • Make remaining model component names architecture-explicit (#728)

     0xShug0

     authored
     901f9edView commit details

     Copy full SHA for 901f9ed

     View code at this pointBrowse repository at this point

   • Make model-local component architecture names explicit (1) (#727)

     0xShug0

     authored
     1716940View commit details

     Copy full SHA for 1716940

     View code at this pointBrowse repository at this point

   • Clarify shared decoder and model architecture names (#725)

     0xShug0

     authored
     eea15b3View commit details

     Copy full SHA for eea15b3

     View code at this pointBrowse repository at this point

   • Rename shared HuBERT encoder as Wav2Vec2 runtime (#724)

     0xShug0

     authored
     342d26dView commit details

     Copy full SHA for 342d26d

     View code at this pointBrowse repository at this point

   • Add standalone Tone Color VC model support (#716)

     Show description for f99bd1e
     0xShug0

     authored
     f99bd1eView commit details

     Copy full SHA for f99bd1e

     View code at this pointBrowse repository at this point
6. Commits on Sep 27, 2026

   • Add native SAM Audio separation (#711)

     Show description for 90c56c2
     0xShug0

     authored
     90c56c2View commit details

     Copy full SHA for 90c56c2

     View code at this pointBrowse repository at this point
7. Commits on Sep 26, 2026

   • Add native SAMSONE audio understanding (#709)

     Show description for 7b1cd06
     0xShug0

     authored
     7b1cd06View commit details

     Copy full SHA for 7b1cd06

     View code at this pointBrowse repository at this point

   • Add GigaAM v3 and multilingual ASR support (#708)

     0xShug0

     authored
     f71fcd8View commit details

     Copy full SHA for f71fcd8

     View code at this pointBrowse repository at this point

   • Add optimized Maya1 TTS support (#707)

     0xShug0

     authored
     ac435fbView commit details

     Copy full SHA for ac435fb

     View code at this pointBrowse repository at this point

   • Add HTDemucs 6-stem source separation (drums, bass, vocals, other, guitar, piano) (#694)

     Show description for 94bd465
     

     SMGoro

     and

     root

     authored
     94bd465View commit details

     Copy full SHA for 94bd465

     View code at this pointBrowse repository at this point

   • Add speaker-tagged transcription to Nemotron ASR (#681)

     Show description for da18f6d
     LysanderdeJong

     authored
     da18f6dView commit details

     Copy full SHA for da18f6d

     View code at this pointBrowse repository at this point
8. Commits on Sep 25, 2026

   • nemotron_3_diar: write speaker probabilities as safetensors (#686)

     Show description for e79205f
     LysanderdeJong

     authored
     e79205fView commit details

     Copy full SHA for e79205f

     View code at this pointBrowse repository at this point

   • nemotron_asr: move options to the model spec (#685)

     Show description for 7c79c9a
     LysanderdeJong

     authored
     7c79c9aView commit details

     Copy full SHA for 7c79c9a

     View code at this pointBrowse repository at this point
9. Commits on Sep 24, 2026

   • Improve Nemotron 3 diarization NeMo parity (#677)

     LysanderdeJong

     authored
     de76accView commit details

     Copy full SHA for de76acc

     View code at this pointBrowse repository at this point

   • pocket_tts: chunk text like the reference, UTF-8 aware text prep (#671)

     Show description for fc24c99
     

     bvisible

     and

     0xShug0

     authored
     fc24c99View commit details

     Copy full SHA for fc24c99

     View code at this pointBrowse repository at this point

   • pocket_tts: use flow_lm.flow.depth for the flow net (fixes *_24l loading) (#670)

     Show description for b780181
     

     bvisible

     and

     0xShug0

     authored
     b780181View commit details

     Copy full SHA for b780181

     View code at this pointBrowse repository at this point

   • higgs_audio_tts: optimize CUDA KV-cache transitions (#669)

     Show description for 267ad4f
     

     mirek190

     and

     0xShug0

     authored
     267ad4fView commit details

     Copy full SHA for 267ad4f

     View code at this pointBrowse repository at this point
10. Commits on Sep 23, 2026

    • Reduce Nemotron-3 diarization graph memory usage

      0xShug0

      committed
      f5a5552View commit details

      Copy full SHA for f5a5552

      View code at this pointBrowse repository at this point

    • Add Nemotron-3-Diarization (preview) support

      Show description for 4b11168
      0xShug0

      authored
      4b11168View commit details

      Copy full SHA for 4b11168

      View code at this pointBrowse repository at this point

    • Add Moonshine ASR audio chunking (#662)

      Faazil-jpg

      authored
      ab26890View commit details

      Copy full SHA for ab26890

      View code at this pointBrowse repository at this point

    • Scope native batching to transcription endpoint

      0xShug0

      committed
      f165a21View commit details

      Copy full SHA for f165a21

      View code at this pointBrowse repository at this point

    • Add Nemotron 3 diarization support

      0xShug0

      committed
      dfb4468View commit details

      Copy full SHA for dfb4468

      View code at this pointBrowse repository at this point
11. Commits on Sep 22, 2026

    • yue2: continue a song from given semantic tokens (#641)

      Show description for d32a6a0
      engival

      authored
      d32a6a0View commit details

      Copy full SHA for d32a6a0

      View code at this pointBrowse repository at this point

    • yue2: tile the eager NAR attention so long songs fit a 4 GiB buffer (#642)

      Show description for 5d786f9
      engival

      authored
      5d786f9View commit details

      Copy full SHA for 5d786f9

      View code at this pointBrowse repository at this point

    • yue2: stop_after ends a run after the score or the semantic stage (#636)

      Show description for 2fff3c8
      engival

      authored
      2fff3c8View commit details

      Copy full SHA for 2fff3c8

      View code at this pointBrowse repository at this point

    • perf(roformer): pipeline CUDA chunk processing (#633)

      Show description for 741fe03
      mirek190

      authored
      741fe03View commit details

      Copy full SHA for 741fe03

      View code at this pointBrowse repository at this point