Course narration TTS API: lesson modules at batch economics — Audra

Batch narration pipelines need predictable MP3 output, not a realtime agent stack. Not for realtime voice agents or cloning.

Who this is for

Teams shipping e-learning / LMS pipelines as mastered MP3/WAV files
Pipelines that already POST text over HTTP or the OpenAI SDK
Builders comparing ElevenLabs list rates for non-cloning workloads
Anyone who needs −16 LUFS without a separate loudness pass

How e-learning / LMS pipelines pipelines usually look

Treat Audra as a file factory: chunk scripts server-side (or send up to 50k sync / 500k via jobs), keep one voice slug for brand consistency, and store mastered MP3s in your CDN. e-learning / LMS pipelines is a batch HTTP workload — not a realtime agent stack.

Stable voice slug across episodes/modules to avoid weekly voice drift
−16 LUFS out of the API — skip a separate loudness pass for most uploads
POST /v2/speech/jobs + webhook_url when a script exceeds sync limits
Model cost on /calc before locking a course or channel budget

Request pattern

curl https://audratalks.com/v2/speech \
  -H "Authorization: Bearer sk-live-YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"audra-essential","text":"Your narration script.","voice":"june","format":"mp3"}' \
  --output out.mp3

OpenAI SDK users set baseURL to https://audratalks.com/v1 and model audra-essential — see /migrate. Next: see the courses use case at /use-cases/courses.

Cost sanity check

Want to hear a sample on your own script? Get an API key at /dashboard — 25k free characters, no card.

FAQ

Should I use sync /v2/speech or async jobs?

Sync for scripts up to 50k characters. Use POST /v2/speech/jobs (up to 500k) plus optional webhook_url for longer modules or episodes.

Do I need ffmpeg for loudness?

Usually no — Audra masters MP3 to −16 LUFS. Spot-check a sample episode before deleting your post-process step.

Does Audra support realtime voice agents?

No. Audra is a batch TTS API for mastered MP3/WAV files. For sub-200ms conversational agents, stay on a realtime-oriented provider.

Can I clone a custom brand voice?

No. Audra ships preset voices only. Use ElevenLabs (or similar) when cloning is a hard requirement — see /vs/elevenlabs.

Next steps

Primary next step: see the courses use case at /use-cases/courses
A/B voices on /compare
Honest matrices: /vs/elevenlabs and /vs/openai-tts
Cost model: /calc
Voice catalog: /voices

Audra is not a clone of ElevenLabs realtime or voice cloning. If that is your core product, stay there. If you need batch narration economics with broadcast loudness, start at /dashboard.