Eleven3

🗣️ Voice Batches

The batch renderer turns validated JSONL text items into cached MP3s with the selected voice through the app proxy.

🎤 Request

  • method: POST.
  • route: /api/elevenlabs/tts?stream=true.
  • voice_id: nsiEp1alHkdiizNqCK2X.
  • model_id: eleven_v3.
  • output_format: mp3_44100_128.
  • voice_settings: { "stability": 0.5, "similarity_boost": 0.75 }.

🔁 Render Loop

  • read batches/wwtbam-de-eleven3-batch.jsonl.
  • parse one JSON object per line.
  • reject raw placeholders before any paid call.
  • render with max two concurrent requests.
  • write each response to a temp file.
  • validate MP3 magic and size.
  • atomically replace the final filename.
  • skip already-valid MP3 files on resume.

📁 Outputs

  • audio_dir: batches/audio/.
  • manifest: batches/wwtbam-de-eleven3-audio-manifest.json.
  • log: batches/audio-generation-log.jsonl.
  • expected_count: 600 MP3 files.
  • observed_valid: 600 MP3 files.
  • failed_renders: 0.

🛡️ Public Boundary

  • keep provider credentials in server/runtime files only.
  • never publish private credentials, session material, direct upstream details, or raw loopback URLs.
  • keep generated audio outside Quartz content unless a future public release explicitly requires publishing it.

0 items under this folder.