Skip to content

v0.15.6

Latest
Latest

Choose a tag to compare

View all tags
@Alex-Wengg Alex-Wengg released this 19 Aug 06:24
· 9 commits to main since this release
v0.15.6
4dbf4f9
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

What's Changed

  • chore(download): remove migration-only characterization scaffolding (#781) #782
  • fix(tts/kokoro-ane): smart-apostrophe contractions + hyphenated lexicon lookups (#774, #775) #783
  • fix(tts): normalize decade and bare-year forms in English text (#776) https://github.com/FluidInference/FluidAudio/pull/780a
  • feat(tts): route English frontends through native NeMo TN when available #784
  • Seam-gap repair pass for long-form chunk-merge content drops by @maboa in #761
  • feat(asr): add language hint to SlidingWindowAsrConfig by @patlux in #786
  • feat(tts/kokoro): byte-exact NeMo text normalization before G2P #790
  • fix(download): preserve model cache on transient network errors by @dtopel in #792
  • feat(tts): LuxTTS backend 48 kHz zero-shot voice cloning (CoreML) #785
  • fix(itn): protect natural-language ambiguous words from native normalization #796
  • fix(tts/pocket-tts): enable voice cloning for 24-layer non-English packs (#793) #797
  • OfflineDiarizerManager: split process() into prepare()/cluster() for cacheable segmentation+embeddings by @ComicBit in #789
  • Remove redundant audio decoding in CLI transcribe path by @hamzaq2000 in #799
  • fix(asr): end-align the final window at the last speech-bearing frame (#747) #800
  • feat(asr): expose non-mutating vocabulary candidate evidence by @CiprianSpiridon in #805
  • fix(tts): make LuxTTS resources signable on iOS by @CiprianSpiridon in #806
  • fix(sortformer): streaming mel emits phantom frames, drifting all timestamps late by @predict-woo in #807
  • feat(download): stall watchdog + DownloadConfig plumbing (#810) #811
  • feat(cli/process): add --compute-units flag for offline diarization #812
  • Expose token timings from the Unified offline batch path by @NylonDiamond in #814
  • fix(tts/kokoro-ane): throw instead of trapping on non-finite PostAlbert durations #816
  • fix(tts/kokoro-ane): warn on BNNS-crash-prone OS builds (26.4–26.5.x) + document both Apple bug classes #818
  • fix(download): include CoreML bundle internals for subPath repos (StyleTTS2 corruptedModel) #821
  • feat(tts/kokoro-ane): Japanese variant + MiniMax benchmark support (--phonemes) #820
  • fix(asr): preserve merge token order at chunk seams (#825) #830
  • fix(download): interrupted first download no longer bricks streaming ASR managers (#819); Unified int8 A16 compat docs (#828) #829
  • fix(asr/eou): debounce on wall-clock silence, not consecutive EOU emissions #831
  • fix(download): stop downloading never-loaded sibling bundles in subPath repos (#826) #832
  • fix: prevent integer overflow in AudioSourceFactory frame count calculation by @jtjones1028 in #813
  • docs(asr): clarify vocabulary alias match semantics by @CiprianSpiridon in #834
  • fix(tts/kokoro-ane): route noise+tail to CPU by default on OS 27 #849
  • fix(tts/kokoro-ane): keep BNNS crash advisory active on iOS 26.6+ #848
  • fix(asr/streaming): temporally gate token dedup to stop prefix-collision loss (#787) #788
  • fix(audio): tolerate the phantom tail of packetized formats in resampleAudioFile by @maboa in #845
  • fix(asr/tdt): scope applyEnglishBlocklist to French only #847
  • feat(tts/kokoro-ane): expose predicted token durations by @Bbrizly in #854
  • Exclude Parakeet benchmark documentation from target by @aindaco1 in #837
  • feat(speaker): CAM++ speaker-embedding backend (CoreML) [beta] #652
  • feat(vad): FSMN-VAD backend (CoreML) [beta] #653
  • feat(asr/canary): Canary-1B-v2 AED engine + CTC-spotter custom vocab [beta] #709
  • feat(tts): add NeuTTS-2E backend (emotional English TTS, Qwen3 + NeuCodec) [beta] #822
  • perf: improve performance by @yanxxl in #804
  • feat(tts): Inflect v2 (Micro/Nano) CoreML backend [beta] #823
  • fix(asr): post-#804 cleanup — format CI, pointer lifetimes, shared argmax, timestamp tail #857
  • Add ModelRegistry.repoOverrides so mirrored repos can be resolved by @omachala in #824
  • style(tests): wrap over-long ModelRegistryTests lines from #824 #859
  • fix(offline-diarizer): pyannote-parity clustering — threshold semantics, constraint count, constrained assignment #802
  • fix(tts): repair legacy Kokoro ANE caches on OS 27 #858
  • fix(asr/streaming): re-decode final window from frame 0 — stops #855 trailing-word loss #861
  • perf(download): skip unused PocketTTS variants + concurrent subdirectory fetches #863
  • feat(asr/unified): vocabulary boosting on the unified Parakeet path #862
  • fix(download): latch pre-attach cancellation so cancelled downloads never start #864

New Contributors

Full Changelog: v0.15.5...v0.15.6

Contributors

maboa, ertra, and 14 other contributors
Assets 2
Loading