Voice greeting on a still photo: TTS first, then H3 Max lip sync

A still photo can speak a greeting if you do it in two calls: generate the voice with TTS, then send the photo and the audio to POST /v1/minimax/h3-max/lip-sync. The audio must be 5 to 14.8 seconds, and TTS costs $0.0475 per 1,000 characters, so a 400-character greeting is $0.019 of voice before the lip-sync charge.

Step 1: the voice

Use POST /v1/tts-1.0/generate with a transcript and a voice selector (avatar_handle or voice.id). Ask for wav and measure the duration of the file you get back, because the lip-sync call needs duration_seconds.

Step 2: the lip-sync call

The body takes exactly one visual source (image_url or an avatar handle) plus a Sume-hosted audio_url and duration_seconds. The default resolution is 768p; 480p and 1080p are also allowed. Sume reserves the list price times 1.25 at admit, so read the live figure in GET /v1/catalog rather than assuming one.

curl -X POST https://api.sume.com/v1/minimax/h3-max/lip-sync \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: greeting-photo-001" \
  -d '{
    "image_url": "https://example.com/portrait.png",
    "audio_url": "https://media.sume.com/artifacts/artf_demo/greeting.wav",
    "duration_seconds": 9.4,
    "resolution": "768p"
  }'

Limits and gotchas

• Audio under 5 seconds or over 14.8 seconds is refused, not clamped. Split a longer script into several sentences and several calls.
• audio_url must be on the Sume media host and no larger than 10 MB; the TTS output already is.
• The still must have an aspect ratio between 0.4 and 2.5.
• Never put someone else's face on a script they did not approve.