Appearance
ElevenLabs
ElevenLabs (/media/elevenlabs) generates natural-sounding text-to-speech voiceovers, still images, and short video clips. Pick a voice and tune stability for voiceovers, or prompt Gemini / GPT Image / Veo models for visuals.
Where to configure
Configure your ElevenLabs API key under Settings → Media. See Settings for provider setup.
Image & Video requires an ElevenLabs Pro plan. The Flows API (image + video generation) rejects keys on lower tiers with
402 paid_plan_required. TTS works on any plan.
Tabs / operations
| Tab | Operation | Key fields | Produces |
|---|---|---|---|
| Text → Speech | Audio / TTS | Text, model, voice, stability, similarity boost, style exaggeration, speaker boost | Voiceover audio |
| Text → Image | Image | Model (Gemini 3.1 Flash Image, Gemini 3 Pro Image), prompt, aspect ratio, resolution | Still image |
| Text → Image (GPT) | Image | Model (GPT Image 2, GPT Image 1), prompt, aspect ratio, quality | Still image |
| Text → Video | Video | Model (Veo 3.1 Fast, Veo 3.1), prompt, duration (4/6/8 s), aspect ratio, resolution, generate audio | Video clip (MP4) |
TTS models offered: Multilingual v2 (default), Eleven v3, Flash v2.5, and Turbo v2.5 (legacy — Flash is recommended over it). The retired eleven_monolingual_v1 / eleven_multilingual_v1 models are no longer listed; ElevenLabs rejects them with unsupported_model.
Generation flow
- TTS is synchronous. The audio is returned inline and appears in the render queue immediately, already saved to
/files. - Images are submitted to the ElevenLabs Flows API and polled inside the request (a few seconds); the finished image lands in
/filesbefore the studio returns. - Videos are queued: the studio returns immediately and the clip appears in the render queue when ElevenLabs finishes — expect minutes, not seconds. No webhook is required; the media-jobs poll drives completion.
From the queue card you can play or preview the result, open it in the Designer timeline, or post it. See Media Studios overview for the shared hand-off flow.
Caveats
- The Image & Video API is Pro-plan gated by ElevenLabs; a
402 paid_plan_requirederror means the workspace needs an upgrade, not a Postmill misconfiguration. - ByteDance models (Seedream / Seedance) are not offered — ElevenLabs disables them by default and requires explicit approval (
model_access_denied). - Each Flows model accepts only its own parameters; the studio only sends the fields valid for the selected model.
- Failed generations report ElevenLabs'
failure_reason(e.g.moderated,model_error) on the job card; failed generations are not charged by ElevenLabs. - Field names are native ElevenLabs params (
voice_id,model_id,voice_settings,aspect_ratio,duration_secs, …) and ride straight into the request body. - The default voice is Rachel; the dropdown includes a curated set of premade voices, but any valid
voice_idfrom your ElevenLabs account works. - ElevenLabs does not support standalone avatar generation here.
Related docs
- Media Studios overview — the shared render-queue and hand-off flow.
- Deepgram — speech-to-text and captions.
- Settings — configuring media providers.
Verified against v1.3.0