Simplify audio post-processing; set pad token; add generation logging 110cbbd nexusbert commited on Nov 10, 2025
Fix smooth generation: add num_beams=5 for local languages, disable early_stopping, fix lang_norm variable, add generation logging 9868a45 nexusbert commited on Nov 8, 2025
Fix audio fade-out: remove unnecessary interpolation when speed_factor=1.0, use scipy.signal.resample for better quality, improve normalization 2a6e73d nexusbert commited on Nov 8, 2025
Fix warnings: add attention_mask support, set pad_token, use do_sample with temperature for /speak and /tts-stream 4af1727 nexusbert commited on Nov 8, 2025
Fix speak-ai: use _transcribe_impl helper to properly handle audio transcription flow feae02a nexusbert commited on Nov 8, 2025
ASR lazy-load: per-repo cache under /tmp; writable caches; disable lifespan; Farmlingua branding ead6187 nexusbert commited on Oct 30, 2025
Reduce storage: move HF caches to /tmp; store model ckpt in /tmp/models; ASR snapshot_download to /tmp with no symlinks b52b307 nexusbert commited on Oct 30, 2025
Docker: remove ASR predownload; rely on runtime snapshot_download 934e4f4 nexusbert commited on Oct 30, 2025
Uvicorn: disable lifespan to avoid CancelledError on startup in Spaces 34b0587 nexusbert commited on Oct 30, 2025
ASR runtime fetch: use snapshot_download in _get_asr with HF_TOKEN; no Docker prefetch required 21e4340 nexusbert commited on Oct 30, 2025
Rename to Farmlingua Speech System; enforce default voices; ignore speaker_name; update /tts-stream to defaults 73082a3 nexusbert commited on Oct 30, 2025
ASR+Ask integration, single /speak, scipy WAV, build-time ASR cache, cleanup af80a9d nexusbert commited on Oct 30, 2025
Fix FileResponse background task; add StreamingResponse cleanup; add sounddevice and system libs b2581bc nexusbert commited on Oct 30, 2025
TTS: use scipy.io.wavfile for WAV output; remove torchcodec dependency path 2dccd23 nexusbert commited on Oct 30, 2025
TTS: robust ckpt handling; download to /tmp then move; make /code writable 055110d nexusbert commited on Oct 30, 2025
requirements: add torchcodec to enable torchaudio save backend d8e7f14 nexusbert commited on Oct 30, 2025
Dockerfile: avoid repo models/ dependency; skip large ckpt; install yarngpt if missing 9499fce nexusbert commited on Oct 30, 2025
Dockerfile: fix HF login RUN block; remove heredoc ; parsing issue 0d4f107 nexusbert commited on Oct 30, 2025