Update README.md

README.md

@@ -1,32 +1,35 @@
 ---
-title: "ActualTTS (Piper CLI, CPU)"
-emoji: "🗣"
-colorFrom: "blue"
-colorTo: "purple"
-sdk: "docker"
+title: ActualTTS (Piper CLI, CPU)
+emoji: 🗣
+colorFrom: blue
+colorTo: purple
+sdk: docker
 pinned: false
 ---
 
 # ActualTTS (Piper CLI, CPU)
 
 FastAPI wrapper around the *Piper* TTS CLI for CPU-only synthesis.  
-Voices are stored under a writable dir and can be provisioned via an API call.
+Voices are cached under a writable dir and can be provisioned via API.
 
 ## Endpoints
 
-- GET /health – shows available voices and directories
-- POST /provision – download/provision a voice  
-  Body (JSON): { "voice": "en_US-amy-medium" }  
-  Known: en_US-lessac-high, en_US-libritts-high, en_US-amy-medium
-- POST /speak – synthesize a WAV and return an audio_url  
-  Body (JSON): { "text": "Hello", "voice": "en_US-lessac-high", "rate_wpm": 170 }
-- GET /file/{name} – fetch a generated WAV
-- WS /ws/tts – near-real-time streaming:  
-  1) send {"event":"init","voice":"en_US-lessac-high","rate_wpm":170}  
-  2) send {"event":"speak","text":"Your paragraph..."}  
-  3) receive {event:"ready", sr, channels}, then *binary* wav bytes and {event:"done"}
+### Health
+GET /health → engine info, available voices, cache dirs
 
-## Health
+### Provision
+POST /provision  
+Body: {"voice":"en_US-amy-medium"}  
+Downloads the ONNX + JSON for the requested voice if missing.
 
-```bash
-curl -s https://<space>.hf.space/health
+### Make WAV (JSON → URL)
+POST /speak  
+Body:
+```json
+{
+  "text": "Hello",
+  "voice": "en_US-lessac-high",
+  "length_scale": 1.08,
+  "noise_scale": 0.35,
+  "noise_w": 0.90
+}