Holiday greeting in 5 languages with H3 Max Lip Sync: $7.50 at 768p

Five holiday greeting videos from one still image, each lip-synced to a different language, cost $7.50 at 768p with MiniMax H3 Max Lip Sync on Sume. At 480p they cost $4.70 and at 1080p $15.00.

Each clip is billed per whole audio second, and the audio window is 5 to 14.8 seconds, which is billed as 15 seconds. So the ceiling for one greeting is 15 × $0.10 = $1.50 at the default 768p.

Rates and the arithmetic

The provider list is $0.05, $0.08 and $0.16 per second for 480p, 768p and 1080p (fal page 2026-09-22). Sume bills list × 1.25, which gives $0.0625, $0.10 and $0.20 per second. Billing is ceil(duration_seconds) times that rate, reserved at admit, captured on completion and refunded on failure.

Constraints that affect planning

The audio must be a Sume-hosted file under 10 MB and between 5 and 14.8 seconds. The API refuses a duration_seconds outside that range rather than clamping it, since the provider silently clips any audio after 14.8 seconds and some speech would be lost. So write each language's greeting to fit about 13 seconds, with a margin.

The visual source is a public HTTPS still or a ready avatar. The aspect ratio of the still must be between 0.4 and 2.5. VEED Fabric 1.0 is the alternative talking-still model, billed at $0.1875 per audio second at 720p, so 15 seconds cost $2.82 against $1.50 here.

Submit one language

Generate the voice with TTS first, then pass its Sume-hosted audio URL to the lip-sync route.

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-es-001" \
  -d '{
    "image_url": "https://example.com/founder.jpg",
    "audio_url": "https://media.sume.com/artifacts/YOUR_ARTIFACT/greeting-es.wav",
    "duration_seconds": 13,
    "resolution": "768p"
  }'

Where to spend and where to save

A greeting on a phone is viewed at about 768p, so 768p is the sensible default and 1080p is for a hero asset only. Run the first language at 480p ($0.94) to check framing and lip shapes, then render the approved still at 768p in all five languages. That sequence costs $0.94 + $7.50 = $8.44, and it avoids paying $1.50 five times for a still that turned out badly cropped.