GitHub/FluidAudio
1
0
Fork 0
mirror of https://github.com/FluidInference/FluidAudio.git synced 2026-06-11 20:24:36 +00:00
Code Issues Packages Projects Releases Wiki Activity

Commit Graph

  • 162e6918ef Revise app logo submission guidelines in README main Brandon Weng 2026-06-11 11:53:12 -04:00
  • 08bbcc9c7b docs: add Hedy to the showcase and featured apps (#682) Julian PscheidandBrandon Weng 2026-06-11 08:52:05 -07:00
  • 336695bdbb Add Parakey to Showcase (#684) rcourtmanandBrandon Weng 2026-06-11 16:49:32 +01:00
  • 8bc713db28 Add TypeWhisper to the FluidAudio showcase (#685) Marco Hillger 2026-06-11 17:48:38 +02:00
  • ab29cead3c feat(asr): expose per-token timings from Nemotron streaming ASR (English + multilingual) (#673) Julian Pscheid 2026-06-10 17:09:36 -07:00
  • f52b60b2b6 feat(asr/eou): opt-in fused decoder+joint_decision path (+7-9% RTFx, WER neutral-or-better) (#680) Alex 2026-06-10 17:02:04 -07:00
  • 01e674a683 fix(kokoro): route the Noise stage to GPU in the M5-safe preset (+~10% synth) (#677) Alex 2026-06-10 16:45:32 -07:00
  • d871ea5660 feat(tts/pocket): ANE placements — rank-4 split-KV models (.ane) + MLState pipeline (.aneState) (#679) Alex 2026-06-10 16:10:46 -07:00
  • cf631747ef fix(download): retry transient per-file failures in downloadRepo (#681) Julian Pscheid 2026-06-10 14:46:54 -07:00
  • 5f9c78febc docs: resolve pyannote segmentation — already GPU-dispatched (plan artifact); new playbook rule: plan-preferred != actual dispatch docs/ane-candidates Alex-Wengg 2026-06-10 11:09:59 -04:00
  • 356d709548 docs: EOU fusion Swift-measured — +7-9% e2e RTFx, WER neutral-or-better, outliers absorbed by production chunking Alex-Wengg 2026-06-10 10:59:38 -04:00
  • 73c1af5a3e docs: EOU fusion gate complete — lean ships (traced fp32-only bit-exactness, identical fp16 behavior) Alex-Wengg 2026-06-10 01:17:08 -04:00
  • 39f46bdca1 docs: EOU campaign round 2 findings (small-graph floor measured, pipeline=0%, lean-vs-traced fusion gate pending) + Nemotron #1 in progress Alex-Wengg 2026-06-10 00:54:34 -04:00
  • 105528fb18 docs(ane): resolve Nemotron ML decoder_joint candidate — 54% ANE is the ceiling (LSTM-gated) Alex-Wengg 2026-06-10 00:54:33 -04:00
  • 84bdc99e71 docs: record Parakeet decode-loop campaign outcome (LSTM gate failed; fusion shipped-ready) Alex-Wengg 2026-06-09 23:42:37 -04:00
  • 7053a6d1ef docs: ANE optimization candidates backlog (post Trials 19-23 + #677) Alex-Wengg 2026-06-09 23:25:09 -04:00
  • 1708125241 chore(asr): remove experimental Qwen3 ASR backend (#676) Alex 2026-06-09 17:12:27 -07:00
  • 2b6564251d Remove experimental Magpie multilingual TTS backend (#674) Alex 2026-06-09 16:40:29 -07:00
  • cbbabf6870 chore(asr): remove experimental Parakeet CTC zh-CN Mandarin model (#675) Alex 2026-06-09 16:40:17 -07:00
  • f3760dc396 fix(kokoro): make the M5-safe routing the default (Kokoro works on M5 out of the box) (#671) Alex 2026-06-08 00:16:23 -04:00
  • b28b1f1f99 docs(supertonic): fix stale maxChunkLengthCJK comment (90→57) (#670) Alex 2026-06-08 00:16:00 -04:00
  • 79d35824ce feat(tts): M5 benchmark re-baseline + Kokoro ane-tail-gpu (M5 fix) + Supertonic int4 default + PocketTTS v2.1 ANE docs (#666) Alex 2026-06-07 19:06:00 -04:00
  • 7f963cdc43 PocketTTS v2.1: fused flow decoder (ANE) + cond prefill + fp16 flowlm (~1.8× RTFx) (#665) v0.15.2 Alex 2026-06-07 08:24:45 -04:00
  • d71efbf471 feat(tts/supertonic3): quantized + ANE-bucketed VectorEstimator (#664) Alex 2026-06-06 19:59:49 -04:00
  • fb2cd01bf7 docs(ane): ANE Profiler residency + latency report (#663) Alex 2026-06-05 17:53:39 -04:00
  • 475885df0e docs: update Models.md to match current model registry (#662) Alex 2026-06-05 16:05:15 -04:00
  • ed66535b69 feat(asr): English Nemotron streaming 2240ms tier + B1 fusion (+50% RTFx, WER-neutral) (#660) v0.15.1 Alex 2026-06-05 12:07:46 -04:00
  • d371275d1d perf(asr): opt-in GPU encoder placement for Parakeet v3 (+~8% RTFx, WER-neutral) (#659) Alex 2026-06-04 23:13:46 -04:00
  • c381018cb0 Add Local Narrator to showcase (#658) Paolo Grimaldi 2026-06-04 20:47:36 -06:00
  • b1c82cfbfd feat(asr): Nemotron 3.5 ASR Streaming Multilingual 0.6B (40 locales, on-device CoreML/ANE) (#657) v0.15.0 Alex 2026-06-04 00:25:36 -04:00
  • a40cc3da2f feat(download): add offline-only enforcement via DownloadUtils.enforceOffline (#632) Lee Crossley 2026-06-02 03:13:45 +01:00
  • 16a039a28c docs(vad): revert Benchmarks.md FSMN section to prior content feat/fsmn-vad-coreml Alex 2026-06-01 10:57:02 -04:00
  • 572178bda3 fix(vad): bound FSMN memory on long audio; mark FSMN-VAD beta Alex 2026-06-01 10:53:42 -04:00
  • 344c5a17ef Merge remote-tracking branch 'origin/main' into feat/fsmn-vad-coreml Alex 2026-05-31 22:43:18 -04:00
  • 82da958110 Merge remote-tracking branch 'origin/main' into feat/campplus-coreml feat/campplus-coreml Alex 2026-05-31 22:38:40 -04:00
  • 2fe5390913 fix(download): fetch repo-root vocab.json for Cohere Transcribe (#649) (#650) Alex 2026-05-31 22:35:35 -04:00
  • fc711d5c7f fix(vad): auto tar-extract VOiCES download + honor --all-files Alex 2026-05-31 22:30:30 -04:00
  • 4a508de43c feat(asr): Paraformer-large (zh) CoreML backend (non-autoregressive, CIF) (#651) Alex 2026-05-31 22:18:14 -04:00
  • 60db11a7ff docs(vad): full MUSAN noise-rejection benchmark (FSMN 81.9% vs silero 69.8% specificity) Alex 2026-05-31 22:07:59 -04:00
  • a97afe5245 feat(vad): FSMN-VAD backend in vad-benchmark + real mini50 results Alex 2026-05-31 21:40:27 -04:00
  • 38d394c889 docs(vad): FSMN-VAD benchmark in Benchmarks.md (frame F1 97.4% vs FunASR, RTFx 1209) Alex 2026-05-31 21:33:35 -04:00
  • 00009b62c2 docs(speaker): CAM++ benchmark in Benchmarks.md (AISHELL EER 0.48%) Alex 2026-05-31 21:32:55 -04:00
  • 838592e85c feat(vad): FSMN-VAD backend + CLI Alex 2026-05-31 21:24:39 -04:00
  • 70fec89a28 feat(speaker): CAM++ speaker-embedding backend + CLI Alex 2026-05-31 21:11:43 -04:00
  • 29efe5a700 feat(asr): SenseVoiceSmall CoreML backend (multilingual, non-autoregressive) (#648) Alex 2026-05-31 19:39:23 -04:00
  • 56607d90b9 Supertonic-3: built-in voice enum + on-demand voice download (#644) v0.14.8 Alex 2026-05-31 00:22:45 -04:00
  • ecce3cca77 PocketTTS: stride-aware conditioning extraction for cloned voices (#643) Alex 2026-05-30 23:48:03 -04:00
  • 8e9fbe4402 TTS: store downloaded models in Application Support, not Caches (iOS) (#642) Alex 2026-05-30 23:37:28 -04:00
  • 3825e75495 feat(nemotron-multilingual bench): optional onset warmup (FLUIDAUDIO_WARMUP_MS) feat/nemotron-onset-warmup Alex-Wengg 2026-05-30 11:37:45 -04:00
  • 6ef369d079 Update Kesha Voice Kit showcase entry (#640) Anton Yakutovich 2026-05-30 16:26:14 +03:00
  • c339ba397d Showcase: add Parleq (#639) Jonathan Yoder 2026-05-29 22:09:50 -04:00
  • 92928c0461 feat(diarizer): expose per-chunk embeddings on DiarizationResult (#633) Robert Marshall Adams 2026-05-29 15:11:35 -07:00
  • 0a7dd2b821 chore(nemotron-multilingual): remove stray MLS results dump from repo root Alex-Wengg 2026-05-29 17:26:26 -04:00
  • 376215a810 fix: address Devin review (argmax dtype, Sendable, warmup) + iOS strict-concurrency Alex-Wengg 2026-05-29 17:24:28 -04:00
  • 6276abe1e4 Merge origin/main into feat/nemotron-multilingual-asr Alex-Wengg 2026-05-29 16:56:36 -04:00
  • 0a5fc31b04 fix(nemotron-multilingual): feed process(samples:) from CLI to satisfy Swift 6.1 Alex-Wengg 2026-05-29 16:54:59 -04:00
  • 7d94a1b18c fix(asr): reduce English drift on French recordings via token blocklist (#630) Matthieu Veinhard 2026-05-29 22:54:22 +02:00
  • d3e5456930 Add timestamp support to EoU Streaming (#629) Kavi GuptaandCopilot Autofix powered by AI 2026-05-29 13:54:06 -07:00
  • ff615561bb ci(nemotron): English B1-vs-K=4 benchmark workflow + benchmark auto-download Alex-Wengg 2026-05-29 16:45:28 -04:00
  • 5b4f72c89a fix(nemotron-multilingual): resolve CI — Swift 6 data race + swift-format Alex-Wengg 2026-05-29 16:33:41 -04:00
  • 80e9fd0932 feat(nemotron-multilingual): make bare decoder/joint optional — support lean B1-only ships Alex-Wengg 2026-05-29 14:29:06 -04:00
  • ae332c1bab feat(nemotron-multilingual): auto-download from HF (mlmodelc-only) Alex-Wengg 2026-05-29 02:22:45 -04:00
  • 0704abaebd fix(cli): expose missing Parakeet langs, add Greek script support (#637) Ryan Patrick Kyle 2026-05-28 17:28:21 -04:00
  • 05d621d682 docs: remove stray leading "d" in README (#636) Roman Passler 2026-05-27 15:50:27 +02:00
  • 5a854cf903 perf(asr/nemotron-multilingual): E7 — B3+B1 noencproj drain priority (NEUTRAL at canonical) Alex-Wengg 2026-05-26 21:58:35 -04:00
  • 1f1b0be9b3 feat(asr/nemotron-multilingual): E4 — smart-spec acceptance-rate counters Alex-Wengg 2026-05-26 21:45:25 -04:00
  • 030ba836c6 perf(asr/nemotron-multilingual): L8 — reuse encProj buffers in smart-spec path Alex-Wengg 2026-05-26 00:23:13 -04:00
  • c129715af5 fix(asr/nemotron-multilingual): propagate smart-spec V2 through shared loader Alex-Wengg 2026-05-25 22:33:32 -04:00
  • 98d7389316 chore(asr/nemotron-multilingual): remove C2 cache INT8 simulation helper Alex-Wengg 2026-05-25 12:50:22 -04:00
  • 6903ceb4bc chore(asr/nemotron-multilingual): C2 cache INT8 simulation helper (env-var gated) Alex-Wengg 2026-05-25 11:15:26 -04:00
  • cb1db16a9a perf+chore(asr/nemotron-multilingual): D cold-start audit, B2-in-drain, vDSP argmax, per-model CU env vars Alex-Wengg 2026-05-24 22:17:08 -04:00
  • 410044d189 Fix/word boost improvements (#634) Sachin DesaiandSachin Desai 2026-05-24 18:09:27 -07:00
  • 2305550c92 perf(asr/nemotron-multilingual): smart-spec default-on with K auto-detect + outputBackings hot path Alex-Wengg 2026-05-24 18:05:37 -04:00
  • 90cb62d773 perf(asr/nemotron-multilingual): smart speculative blank with Swift encoder_proj + B1 drain Alex-Wengg 2026-05-24 15:24:28 -04:00
  • fee6ea67f1 feat(asr/nemotron-multilingual): smart speculative-blank decode path (env-gated, ANE blocker) Alex-Wengg 2026-05-24 14:44:57 -04:00
  • 5406d4be4d perf(asr/nemotron-multilingual): hybrid native-LSTM + CoreML-joint path (negative result) Alex-Wengg 2026-05-24 11:44:38 -04:00
  • 92536cdcf4 perf(asr/nemotron-multilingual): preproc bufs prealloc + VAD hangover + process(samples:) API Alex-Wengg 2026-05-24 11:31:35 -04:00
  • 3dba3aebd4 feat(cli): add WER to multi-stream bench for shared-model correctness check Alex-Wengg 2026-05-24 10:06:07 -04:00
  • 5382f96c7b feat(asr/nemotron-multilingual): shared-MLModel multi-stream Alex-Wengg 2026-05-24 09:55:46 -04:00
  • 0ea3bf48c1 perf(asr/nemotron-multilingual): prealloc decode buffers + VAD + multi-stream Alex-Wengg 2026-05-24 03:11:31 -04:00
  • 1d905f48ad perf(asr/nemotron-multilingual): native RNN-T + speculative-batched + warmup paths (all env-gated) Alex-Wengg 2026-05-24 02:34:06 -04:00
  • 97052e902c feat(asr/nemotron-multilingual): add Earnings22 dataset to multilingual benchmark Alex-Wengg 2026-05-23 22:41:19 -04:00
  • 0cbd20e1cf debug: add FLUIDAUDIO_DISABLE_TRIPLE_STAGE env var to bypass triple-stage pipelining Alex-Wengg 2026-05-23 22:21:04 -04:00
  • dc4e411168 feat(asr/nemotron-multilingual): multi-dataset benchmark — MCV-17, MLS, LibriSpeech Alex-Wengg 2026-05-23 20:35:42 -04:00
  • c79ac5fcd9 feat(asr/nemotron-multilingual): block-streamed long-file read + MLOptimizationHints regression note Alex-Wengg 2026-05-23 20:35:21 -04:00
  • 29c0aee474 perf(asr/nemotron-multilingual): triple-stage pipelining + B1/B2 fusion + output backings + profiling Alex-Wengg 2026-05-23 20:35:07 -04:00
  • 20c55df07f Mention Wyoming/HA feature for macos-speech-server. (#631) Mathijs de Bruin 2026-05-22 18:26:04 +01:00
  • 5bb3a58a1a docs(asr): update Nemotron multilingual numbers with ITN-aware scoring Alex-Wengg 2026-05-19 01:10:04 -04:00
  • 0d7295dd32 feat(asr): language-aware ITN for Nemotron multilingual FLEURS scoring Alex-Wengg 2026-05-18 20:52:00 -04:00
  • ed6253463e feat(asr): Nemotron Speech Streaming Multilingual 0.6B Alex-Wengg 2026-05-17 14:32:18 -04:00
  • 50aa07193e Revise showcase entries in README (#628) Alex 2026-05-19 01:55:13 -04:00
  • 498367c4d7 docs: add iPhone 17 Pro ANE TTS demos to Video Demos (#627) Alex 2026-05-19 01:53:40 -04:00
  • 66baf2112d docs(asr): document parallel chunk processing in LongTranscription.md (#626) Alex 2026-05-18 23:28:51 -04:00
  • 1b122018c4 fix(asr): honor caller modelNames in DownloadUtils for non-baseline files (#524) (#625) Alex 2026-05-18 23:18:32 -04:00
  • 8048812869 fix(asr): add opt-in v3 no-mel decode arbitration (#604) v0.14.7 vdt4534andAlex-Wengg 2026-05-19 11:16:12 +09:00
  • 9ffa26b3c9 feat(cli): expose all unexposed config fields across 6 commands (#616) Bazhen Rzheutskii 2026-05-18 01:04:32 +04:00
  • 3b72d67c23 feat(tts/pockettts): add 5 native-language voices + slim language-pack downloads ~40% (#624) Alex 2026-05-17 14:55:31 -04:00
  • 816a65334e docs: add Dettivo to showcase (#618) Gordon Mickel 2026-05-17 16:07:14 +02:00
  • 364d839311 feat(tts): Supertonic-3 multilingual CoreML TTS — CLI, benchmark, 10-lang results (#617) v0.14.6 Alex 2026-05-17 02:25:49 -04:00
  • d11ef65fd0 Diarization progress (#615) Nick Burns 2026-05-16 23:16:30 -07:00