18-second H3 Max Recast person swap: $6.75 at 768p, $10.13 at 1080p
An 18-second source video through H3 Max Recast on Sume costs $6.75 at 768p and about $10.13 at 1080p. Photo limits, the 15 s shot rule, and a request.

Swapping the people in an 18-second video with H3 Max Recast (h3-max-recast) on Sume costs $6.75 at the default 768p and about $10.13 at 1080p. The price is the source video length, rounded up to whole seconds, times the fal list of $0.30 or $0.45 per second (fal 2026-10-01), times Sume's 1.25 margin.
Recast keeps the source's motion, camera, cuts, and sound, and replaces each person with one you supply as a photo. It does not generate a new scene.
The math for 18 seconds
The duration is the length of the source video, so you do not choose it. Sume bills ceil of the source length. Eighteen seconds is exactly 18.
| Resolution | List per second | Sume per second | 18 seconds |
|---|---|---|---|
| 768p (default) | $0.30 | $0.375 | $6.75 |
| 1080p | $0.45 | $0.5625 | $10.125 (about $10.13) |
What the route accepts
An 18-second source with one 10-second shot and one 8-second shot is fine. An 18-second single take is not, because of the 15-second shot limit. Cut such a clip before you send it.
- One source video as
video_url, 5 to 30 seconds, with no single shot longer than 15 seconds. - One to four photos in
reference_image_urls, one for each new person. - An optional prompt of up to 2,000 characters.
- No text-only generation, and no
aspect_ratio: the source sets the frame.
A request
Send the job through Video Router with an idempotency key, so a retry after a timeout does not create a second job.
curl -X POST https://api.sume.com/v1/video-router/generate \
-H "Authorization: Bearer $SUME_API_KEY" \
-H "Content-Type: application/json" \
-H "Idempotency-Key: recast-18s-001" \
-d '{"model": "h3-max-recast", "video_url": "https://example.com/source.mp4", "reference_image_urls": ["https://example.com/new-person.png"], "resolution": "768p", "mode": "async"}'Draft at 768p, ship at 1080p
The 1080p upgrade adds $3.375 on this clip. Run the 768p version first, check that the swap holds on every cut, and re-run at 1080p only for the final. A failed job is refunded; a completed job with a face you dislike is not.
If the source is only a face to replace and the clip is 4 to 15 seconds with usable audio, the beta Avatar Face Swap is another route, compared in face swap against H3 Max Recast. MiniMax's H3 announcement lists video, image, and audio references as inputs to the H3 model (read 2026-10-07); Recast is a separate Sume row with its own rules above.
Sources
Related posts
More in Pricing
- $200 buys 72 fifteen-second Standard avatar clips on Sume
A 15-second Standard avatar clip with no product is $2.76 on Sume. $200 buys 72 of them ($198.72), with $1.28 left over.
- 25 Kling motion control clips of 8 seconds: $31.50 and the Pro queue
Twenty-five 8-second Kling 3.0 motion control jobs cost $31.50 on Sume. The Pro plan accepts 24 at once, so the 25th needs a second wave. The math.
- 25-second avatar clip with a product: $4.85, $6.45 or $14.50 by tier
On Sume, a 25-second avatar clip with a product costs $4.85 on Standard, $6.45 on Plus and $14.50 on Max. Per second: $0.194, $0.258 and $0.58.
- $250 holiday ad budget: split between clips, stills, voice, music
A $250 wallet covers 80 Wan 3.0 ads of 15 seconds, 20 retakes and 10 winners re-rendered at 1080p, plus stills, voice, music and renders: $244.14 on Sume.
Written by Sume