feat: replace edge-tts with piper-tts for Off the Grid badge
carlosduplar
6f92dc6
- Switch from Microsoft Edge cloud TTS (edge-tts) to local piper-tts
- Pre-download fr_FR-siwis-medium voice model (~50MB) on first load
- Update audio pipeline: .wav instead of .mp3, media_type change
- Update README architecture + badge eligibility
Root cause: Gradio 5.0.x gradio_client/utils.py:882 does
'if "const" in schema' where schema can be a bare bool from
Pydantic 1 serialization triggered by old huggingface-hub<0.26.0.
Upgrade to 5.50.0 and drop the stale pin.
- requirements.txt: gradio>=5.0,<6.0 → gradio==5.50.0; remove huggingface-hub<0.26.0
- space_README.md: sdk_version: 5.0.2 → 5.50.0
- app.py: no changes needed (already uses sources=["microphone"] new API)
Pin huggingface-hub<0.26 (HfFolder removed in newer versions, breaks Gradio 5.x)
carlosduplar
002f1f3
Remove pyaudioop (does not exist on PyPI), python 3.12 already has audioop stdlib
carlosduplar
a74a11b
Pin python_version to 3.12 (audioop removed in 3.13), add pyaudioop fallback
carlosduplar
c3a0314
Fix edge-tts version constraint (latest is 7.2.8)
carlosduplar
cfdd7e0
Switch TTS from Mistral API to edge-tts (free, neural French voice)
