tts_local operator: text→speech via Kokoro-82M on MLX (character voice templates)

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

server/operators/tts_local/manifest.json

@@ -0,0 +1,19 @@
+{
+  "id": "tts_local",
+  "name": "Speak (local, Kokoro)",
+  "category": "audio",
+  "description": "Text → speech WAV entirely on this Mac via Kokoro-82M on MLX (Apache 2.0, ungated, ~350MB first download). ~50 built-in voices spanning male/female US/UK = instant character voice templates. Faster than realtime, no API cost, no cloud.",
+  "accepts": [],
+  "produces": ["audio"],
+  "resources": "gpu",
+  "entry": "run.py",
+  "python": "venvs/kokoro/bin/python",
+  "params_schema": {
+    "type": "object",
+    "properties": {
+      "text": {"type": "string", "default": "", "description": "The line to speak"},
+      "voice": {"type": "string", "default": "af_heart", "description": "Kokoro voice id (af_/am_ = US female/male, bf_/bm_ = UK): af_heart, af_bella, af_nicole, am_adam, am_michael, am_onyx, bf_emma, bf_isabella, bm_george, bm_lewis, ..."},
+      "speed": {"type": "number", "default": 1.0, "minimum": 0.5, "maximum": 2.0, "description": "Speaking rate"}
+    }
+  }
+}

server/operators/tts_local/run.py

@@ -0,0 +1,43 @@
+import argparse
+import json
+import subprocess
+import sys
+import time
+from pathlib import Path
+
+ap = argparse.ArgumentParser()
+ap.add_argument("--input", action="append", default=[])
+ap.add_argument("--outdir", required=True)
+ap.add_argument("--params", default="{}")
+a = ap.parse_args()
+p = json.loads(a.params)
+
+text = (p.get("text") or "").strip()
+if not text:
+    print("ERROR: text is required")
+    sys.exit(1)
+
+voice = p.get("voice", "af_heart")
+speed = float(p.get("speed", 1.0))
+outdir = Path(a.outdir)
+
+cmd = [
+    sys.executable, "-m", "mlx_audio.tts.generate",
+    "--model", "prince-canuma/Kokoro-82M",
+    "--text", text,
+    "--voice", voice,
+    "--speed", str(speed),
+    "--file_prefix", str(outdir / f"tts_{voice}"),
+]
+t0 = time.time()
+r = subprocess.run(cmd, capture_output=True, text=True)
+if r.stdout:
+    print(r.stdout[-2000:])
+if r.stderr:
+    print(r.stderr[-2000:], file=sys.stderr)
+
+wavs = sorted(outdir.glob("*.wav"))
+if r.returncode != 0 or not wavs:
+    print("ERROR: kokoro generation failed")
+    sys.exit(1)
+print(f"done in {time.time() - t0:.1f}s: {', '.join(w.name for w in wavs)}")