ElevenLabs (TTS)

Premium neural text-to-speech. Used by the Generate Voice, Dubbing, and Live Translation pages when you pick ElevenLabs as the TTS method.

Get an API key

1. Sign up at https://elevenlabs.io
2. Open https://elevenlabs.io/app/settings/api-keys
3. Click + Create New Key, name it (e.g. "ai-translate"), copy the key (looks like sk_...)

The free tier gives you ~10,000 characters / month, enough to test. Production usage starts around $5/month.

Configure in the app

In Settings → Service:

1. Paste the key into ElevenLabs API key → Save
2. Enter your preferred Voice ID in Voice ID (find IDs at https://elevenlabs.io/app/voice-lab; copy the ID from a voice's URL). Leave blank for ElevenLabs to pick a default.

In Settings → Voice:

1. Set TTS method to ElevenLabs

2. Pick the ElevenLabs model:

   Model	Best for
   eleven_multilingual_v2 (default)	General use, balanced latency/quality
   eleven_v3	Highest quality (use for production dubs)
   eleven_flash_v2_5	Lowest latency (use for Live Translation)

What it powers

Page	Use ElevenLabs when
Generate Voice	You want premium-quality voiceovers from subtitle files
Dubbing	You want a high-quality dub track on a translated video
Live Translation	You want spoken playback of translated captions in real time

Voice cloning

ElevenLabs supports custom voice cloning (paid plan). Once you've cloned a voice on the ElevenLabs site, paste its Voice ID into Settings → Service → Voice ID and the dubbing / voice-generation pipeline will use it.

Caveats

Pre-flight check

The Voice / Dubbing pages check that your ElevenLabs API key is set before starting work. If it's missing you'll get a friendly dialog pointing you to Settings, not a half-run task.

Live mode falls back automatically

On the Live Translation page, if you've selected ElevenLabs but haven't configured a key, the app falls back to Edge TTS (free) and announces the fallback in the status label so you can fix it when convenient.

FFmpeg still required

ElevenLabs returns audio bytes; the app still uses FFmpeg to convert between formats and to combine timed clips into one file. See FFmpeg setup.

Common errors

Error	Likely cause
AUTH_ERROR	Wrong / expired API key. Re-paste in Settings → Service.
QUOTA_ERROR	Free-tier character limit hit, or paid plan exhausted.
MODEL_NOT_FOUND	The selected ElevenLabs model is no longer available; pick another in Settings → Voice.