v0.7.4
What's Changed
- 🐛 fix hip build error by @Orion-zhen in #486
- refactor: build Fish HIP fast sampler as a proper HIP CMake target by @IIIIIllllIIIIIlllll in #490
- perf(mira-tts): accelerate CPU and Vulkan inference by @mirek190 in #491
- Add native GTCRN audio utility by @0xShug0 in #492
- sanotts: eleven more language voices (cs, de, es, fr, it, pt, ro, ru, tr, ne, hi) and the acoustic calibration adapter by @voidash in #495
- Add VibeVoice ASR Streaming 7B by @0xShug0 in #497
- Add a one-click Colab WebUI with a CUDA T4 release build by @DrewThomasson in #501
- feat(community_models): add Sortformer v2.1 streaming diarization by @LysanderdeJong in #493
- fix(warmbench): voxtral_realtime branch reads an undefined
mode, so every run raises NameError by @Anai-Guo in #506 - Apply upstream Vulkan FILL dispatch fix by @0xShug0 in #508
- Shared eSpeak-ng phonemizer for model frontends by @mirek190 in #502
- perf(fish_audio): keep Fast-AR inputs on HIP stream by @reezex0-ux in #500
- Fix Windows CUDA build when PATH has a stray quote by @CaptainArni in #511
- feat(kokoro): add multilingual GGUF synthesis and optimized CPU inference by @mirek190 in #496
- Use shared eSpeak frontend for Kokoro English by @0xShug0 in #515
- Add Moonshine STT model by @0xShug0 in #510
- Rename Moonshine STT to ASR by @0xShug0 in #518
- fix(vocoders): fix narrowing conversion in HiFT vocoder on 32-bit tar… by @weidu-tpvision in #512
- perf(vibevoice_asr_streaming): flash suffix attention and bounded history window by @LysanderdeJong in #526
- Add Orukeet r3 weight variant to parakeet_tdt by @LysanderdeJong in #523
- fix(pocket-tts): make reference audio fields visible again by @SmilingWolf in #528
- Add portable Linux release artifacts by @0xShug0 in #529
- fix(models): prevent size_t overflow on 32-bit targets in model confi… by @weidu-tpvision in #513
- Add Moonshine ASR GGUF docs and UI entries by @0xShug0 in #531
New Contributors
- @DrewThomasson made their first contribution in #501
- @Anai-Guo made their first contribution in #506
- @weidu-tpvision made their first contribution in #512
- @SmilingWolf made their first contribution in #528
Full Changelog: v0.7.3...v0.7.4