H3 Max Recast: what a 30-second people swap costs at 768p and 1080p

H3 Max Recast (h3-max-recast) replaces the main people in a source video with the people in one to four reference photos, and a 30-second source costs $11.25 at 768p or $16.875 at 1080p on Sume. The rates are $0.375 and $0.5625 per second, which is the provider list of $0.30 and $0.45 times 1.25, as of 2026-10-08.

The output keeps the source length, motion, camera, cuts and soundtrack, so the price is set by the length of the video you upload.

What you send

On POST /v1/videos, send input_references with exactly one video_url and one to four image_url entries, one photo per new person, matched left to right by default. On the Video Router, send video_url and reference_image_urls. The prompt is optional (up to 2,000 characters) and can say who becomes whom or what else to keep or change.

Set duration to the inspected length of the source in seconds, rounded up. The source must be 5 to 30 seconds with no single shot longer than 15 seconds.

curl -X POST https://api.sume.com/v1/video-router/generate \
  -H "Authorization: Bearer $SUME_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"h3-max-recast","video_url":"https://example.com/source.mp4","reference_image_urls":["https://example.com/new-person.png"],"resolution":"768p","duration":30}'

What is not accepted

The default matters for cost: leaving resolution out gives the cheaper tier, a difference of $0.1875 per second.

• No text-only generation: a source video and photos are required.
• No aspect_ratio, frame_images, generate_audio, bitrate_mode, seed or audio references.
• Resolution is 768p or 1080p. Sume defaults to 768p, even though the provider default is 1080p.

Cost against the alternatives

A 15-second recast at 768p costs $5.625. Generating a fresh 15-second clip on MiniMax H3 Max at 768p costs $1.50, and Seedance 2.5 at 720p costs $8.667. Recast costs more per second because it keeps an existing performance, camera and sound, which a fresh generation cannot match. Use it when the footage is already approved and only the people must change.

Planning a batch

Recast is priced by the source length, so a batch of ten 30-second ads costs 10 x $11.25 = $112.50 at 768p, or 10 x $16.875 = $168.75 at 1080p. Test one swap at 768p first. If the result is what you want, rerun the approved clips at 1080p: the extra cost is 30 x $0.1875 = $5.625 per clip.

Sume also lists Genjutsu Motion Transfer for the opposite job: one source video plus 1 to 8 reference images, at 480p or 720p, for 4 to 30 seconds. The two are different tools; Recast replaces the people, while Genjutsu transfers motion onto a new subject.

Checklist

• Measure the source length first and round up.
• Use clear, front-facing photos, one per person.
• Start at 768p and rerun at 1080p only on the approved cut.
• Send an Idempotency-Key so retries return the original job.