The open-source AI voice studio. Clone, dictate, create.
Commits on main
fix(docker): fix startup crash and permission errors in the container
ab9a197 harry committed 2 months ago
Merge pull request #1130 from jamiepine/prep/pr-1112
fee218e capy-ai-staging[bot] committed 1 month ago
Merge pull request #1144 from jamiepine/prep/pr-1031
39b1c91 capy-ai-staging[bot] committed 1 month ago
Merge pull request #1146 from jamiepine/prep/pr-659
3221b27 capy-ai-staging[bot] committed 1 month ago
fix(build): keep actool failure fatal for release builds; surface its full output
370fd15 jamiepine committed 1 month ago
Make actool failure non-fatal in build.rs — allows building without full Xcode
c31703a Will Anderson committed 6 months ago
fix(mlx): fold Qwen3 LLM load+generate into one worker submission; bind model locally in generate closures
dd8ab5c jamiepine committed 1 month ago
fix(mlx): keep unload_model off the MLX worker so it cannot stall the event loop
cf98488 jamiepine committed 1 month ago
fix: extend MLX thread-affinity fix to the Qwen3 LLM backend
2693c98 Seunghun Lee committed 4 months ago
fix(mlx): fold reload+inference into one MLX-worker submission
57ef663 Ron David Ben Ishay committed 3 months ago
fix(mlx): address review feedback on the thread-affinity fix
fa8db82 Ron David Ben Ishay committed 3 months ago
fix(mlx): pin all MLX ops to a single worker thread
f135a47 Ron David Ben Ishay committed 3 months ago
fix(docker): adopt the owner of a host-mounted /app/data too, never chown host dirs
76dd300 boessu committed 3 months ago
Address review: targeted chown, fail-fast on non-writable dirs, drop stale HF cache mount
6f66f93 LaurinceG4N0 committed 2 months ago
Fix: writable data dir, HF cache on data volume, and install SoX
1d8d324 LaurinceG4N0 committed 2 months ago
fix(docker): chown ~/.cache so the app user can write caches
49b55fc Thiago Margatho committed 4 months ago
fix: Dockerfile HuggingFace cache and generations directory permissions
b39e5b2 virtuallyunknown committed 3 months ago
refactor(stt): name the Whisper sample rate and window constants
86dc46b jamiepine committed 1 month ago
fix(backend): keep the pad-to-30s Whisper path for clips under 30s
51882b5 jamiepine committed 1 month ago
fix(backend): enable long-form Whisper transcription on PyTorch path
e61c853 nox committed 6 months ago
fix(kokoro): trim trailing silence and run-on noise on short prompt synthesis (#960)
615aeae devangkantharia committed 3 months ago
fix(generation): make post-generation cleanup best-effort and cover /generate/stream
b788dc3 jamiepine committed 1 month ago
refactor(generation): simplify post-generation cache cleanup
17fd1dd jamiepine committed 1 month ago
fix(backend): prevent unbounded memory accumulation over consecutive TTS generations (#923)
38389db devangkantharia committed 3 months ago
fix(tests): repair test_profile_duplicate_names so the suite can run
63f0945 Lvigentini committed 3 months ago