Spaces:
Running on Zero
Running on Zero
Commit History
-
Add example text: 我要吃蚵仔煎然後去丟垃圾。
Verified signature24dd159 -
Add speech-rate slider (pitch-preserving time-stretch post-processing, shared across tabs)
Verified signature2d419a6
-
Add adjustable cfg_value / inference_timesteps sliders (shared across tabs, defaults = model-card recommendation)
Verified signaturea8231fb -
Voice clone: single reference-clip mode only (drop centroid/ECAPA stable mode + speechbrain)
Verified signature355336d -
Wording: neutral phrasing for reference-clip support boundary
Verified signaturedc2d2f8 -
Update for checkpoint step_0006000
Verified signature23e9d67 -
Update for checkpoint step_0006000
Verified signaturec5dba79 -
step_0006000: fast transcript-free clone (reference_wav_path), new default params (cfg 2.0/steps 10)
Verified signaturee66cd92
-
Add speaker selection (李宏毅 + 女聲) via multi-speaker centroids
Verified signature7b111a1
-
Voice clone: read 10 sentences -> average ECAPA over all clips for a stabler centroid
Verified signature9a401f5 -
Voice cloning: extract ECAPA speaker centroid from upload (no transcript); add speechbrain
Verified signature5fc2aef -
Fix voice cloning: use prompt_wav+prompt_text (reference_wav_path produced gibberish); remove debug
Verified signature5e31159 -
TEMP: add /tts_debug endpoint to diagnose voice-clone content failure
Verified signature7afaa46 -
Long-text streaming: stream until GPU time budget (~100s) instead of fixed 12-sentence cap
Verified signaturea858c6e -
Streaming tab: sentence-by-sentence streaming for long text (synth & play per sentence)
Verified signature01153e3 -
Streaming tab: collect full audio then play smoothly (ZeroGPU is sub-realtime)
Verified signaturedd9fd54 -
Recreate Space (zero-a10g) after region corruption; streaming included
Verified signature927c1a5 -
initial commit
Verified signature96187a5