Make a Talking Photo Speak Spanish: TTS Language, Then Fabric

To make a still photo speak Spanish on Sume, generate the Spanish audio first with the text-to-speech language set to es, then send that audio and the still to Fabric. Avatar Video itself should not be used for Spanish: Sume's guidance is that scripts there should be English, because non-English speech breaks caption alignment.

Prices are from API pricing, read on 2026-10-05: text to speech is $0.0475 per 1,000 characters, and Fabric is $0.1875 per second at 720p.

What are the two steps?

Step one, text to speech: send the Spanish transcript, a voice and language. If the voice's language and language disagree, the API can refuse the request with a 409 tts_voice_language_mismatch before any charge, and you can only continue by setting confirm_language_mismatch. Pick a Spanish voice instead of overriding.

Step two, Fabric: send the generated audio as audio_url (it must be Sume-hosted, 10 MB at most), the measured duration_seconds, and one visual source. The preflight checks only host and size, not the language, so a mismatched voice is yours to catch by listening.

POST https://api.sume.com/v1/veed/fabric-1.0
{
  "image_url": "https://media.sume.com/portrait.png",
  "audio_url": "https://media.sume.com/saludo-es.mp3",
  "duration_seconds": 30,
  "resolution": "720p"
}

What does a 30-second clip cost?

About 75 Spanish words is roughly 450 characters, which is $0.02 of speech at $0.0475 per 1,000 characters. The Fabric part dominates: 30 seconds at 720p is $5.63 (30 x $0.1875 = $5.625), or $3.00 at 480p.

What does this not do?

It does not translate: you supply the Spanish transcript. It does not converse; the still moves to the audio you gave it. For Avatar Video scenes in other languages, see the limits in Generate avatar video and keep spoken scripts English.