All pull requests

Pull requests

• Add a script to use multilingual PocketTTS models from KevinAHM/pocket-tts-onnx

  #4148·haex-space opened Oct 30, 2026·

  ·
  3

• Fix VAD Flush dropping samples that wait for a full window

  #4147·ANDROIDHASSAN opened Oct 30, 2026Contributor·

  ·
  2

• Add English number and symbol rule FSTs for Kokoro

  #4144·JulianPscheid opened Oct 30, 2026Contributor·

  ·
  8

• Expose Whisper language ID probability distribution on OfflineRecognitionResult

  #4143·matiaslin opened Oct 29, 2026Contributor·

  ·
  1

  21

• Add multilingual Nemotron Labs Unified ASR streaming support

  #4141·JulianPscheid opened Oct 29, 2026Contributor·

  ·
  1

  5

• Add Lithuanian Piper voice lt_LT-ingute-medium

  #4140·RobertasTa opened Oct 29, 2026·

  ·
  5

• Fix buffered streaming for NeMo Parakeet Unified

  #4133·JulianPscheid opened Oct 29, 2026Contributor·

  ·
  12

• Add Nemotron-3-Diarization (Sortformer) speaker diarization — rebased #4032

  #4119·Flow-Mind-Company opened Oct 27, 2026·

  ·
  1

  7

• Accept hotwords given as tokens from tokens.txt

  #4088·Hugo0 opened Oct 23, 2026Contributor·

  ·
  3

• Add Uzbek GigaAM model documentation and example

  #4072·muhammadjonsaidov opened Oct 20, 2026·

  ·
  15

• Add Yap to 'Projects using sherpa-onnx'

  #4066·yumaheymans opened Oct 19, 2026·

  ·
  2

• Add AX650 ZipVoice TTS backend

  #4038·Abandon-ht opened Oct 10, 2026Contributor·

  ·
  4

• qwen3-asr: accept the vocab.json file path as the tokenizer argument

  #4037·paoloantinori opened Oct 10, 2026Contributor·

  ·
  1

  8

• qwen3-asr: log the max_total_len generation break; raise default 512 -> 1024

  #4035·paoloantinori opened Oct 10, 2026Contributor·

  ·
  1

  1

• Add Outmake to 'Projects using sherpa-onnx'

  #4033·DaneBentley opened Oct 10, 2026·

  ·
  2

• Add Nemotron-3-Diarization (Sortformer) speaker diarization

  #4032·silviot opened Oct 10, 2026·3 tasks·

  ·
  1

  14

• Batch pyannote segmentation windows in offline speaker diarization

  #4021·Eoin-Houstoun opened Oct 5, 2026Contributor·

  ·
  1

  3

• Fix sherpa-onnx-cli text2token attaching extra info to the wrong line

  #4008·MohammadHijjawi97 opened Sep 30, 2026Contributor·

  ·
  1

• Fix overflows, int32 sign and chunk handling in the WAV reader

  #4007·antonsoo opened Sep 30, 2026Contributor·

  ·
  6

• Reject an integer that does not fit in int64

  #4000·SashaMIT opened Sep 28, 2026·1 task done·

  ·
  3

• Fix keyword acoustic-threshold check to use the matched tokens

  #3998·kyo-zzz opened Sep 27, 2026Contributor·

  ·
  1

• rust: expose ys_probs in the online RecognizerResult

  #3996·voicetel opened Sep 27, 2026·

  ·
  3

• kws: keep encoder states in the automatic silence reset (fixes #3990)

  #3994·dime-online opened Sep 27, 2026Contributor·

  ·
  1

  2

• Add Dolphin attention decoder with forced language/region

  #3986·losewayy opened Sep 25, 2026Contributor·4 tasks done·

  ·
  8

• add word-level timestamps for qwen3-asr via forced aligner (#3552)

  #3984·losewayy opened Sep 25, 2026Contributor·3 of 4 tasks·

  ·
  1

  11