German voice-over on Sume: TTS language field plus cue captions

A 300-character German promo costs about $0.31 on Sume: TTS ($0.0143), one Timeline render ($0.10) and one caption job with your own German cues ($0.20). The TTS request has a language field for non-English text, while Avatar 1.0 stays English-only.

Localization on Sume has a clear line. Voice-over, captions and Timeline are language-aware on the request. Talking avatars are not, and voice cloning is an app-only feature, so this is a recipe for a voice-over over stills or clips. Rates come from the Sume pricing code (read 2026-10-10).

What is the endpoint chain?

The pattern is the same as an English voice-over with two changes: the language field and your own translated cues.

• Translate the script yourself or with your own tool; the TTS endpoint reads the text you send.
• POST /v1/tts-1.0/generate with the German text, language: "de" and timestamps: {words: true}. Confirm the accepted codes in the request schema on the API reference.
• POST /v1/timeline-1.0/render with the TTS URL as audio.url and your existing stills or clips. The video can be the same file you used for the English version.
• POST /v1/video-captions with cues in German, or words from the TTS result. Check that the caption style you pick has the glyphs your text needs.

What does it cost per language?

The visuals are reused, so each added language is voice, render and captions only.

What are the limits?

No avatar localization: a talking avatar is generated from an English script only. If you need a presenter in German, use stills or b-roll with a voice-over instead. Voice cloning is not available through the API; pick a catalogue voice and test it, because a voice that sounds natural in English may sound flat in another language.

German text runs longer than English, often by a quarter or more. Set the spine from the TTS duration_seconds, not from the English timing, and use generation_config.speed (0.6 to 1.5) when a line must fit. Burned-in captions are separate from voice, so recaption instead of regenerating when you fix a word. See Video captions for style and language handling.

Machine translation can be wrong in places a native speaker would catch. Have a person review any ad copy before posting.

How does this compare to an English-only avatar post?

The English avatar path costs more per second ($0.184 on standard); the voice-over path costs under five cents per 1,000 characters. For multi-language reach, a voice-over over a shared visual is the cheaper route. See AI avatar for event videos for the English-only side and the voice-over post for the render.

Pre-publish checklist

Have a native speaker read the script and listen once to the voice. A small stress or number-reading problem is easy to fix by editing the script and re-running the TTS job.

Because the visuals are shared, a fix only re-runs voice, render and captions, which is about $0.31 for 300 characters.

• Spell numbers the way they should be spoken.
• Check line breaks in cues on a phone.
• Keep one file per language, named by code.