Video-to-video AI cost per clip: Recast and Genjutsu

Editing an existing clip costs several times more per second than most text-to-video rows. On Sume, H3 Max Recast bills $0.375 per second at 768p and $0.5625 at 1080p, and Higgsfield Genjutsu bills $0.3975 per second at 480p and $0.85125 at 720p. A 10-second Recast clip at 768p costs $3.75.

Cost per source length

The charge follows the input clip: the duration you are billed for equals the source length, rounded up to a whole second. Recast accepts 5 to 30 seconds and Genjutsu 4 to 30 seconds.

Plan before you upload

Trim first. Every second of input costs the same whether or not the edit touches it, so a 30-second source for a 6-second change costs five times what a trimmed copy would.

• Trim to the span you want changed, then run the edit.
• Test at 480p on Genjutsu before 720p: the rate roughly halves.
• Use 768p on Recast for review, 1080p only for the final.

A worked example

You have a 20-second product clip and want a different presenter in the first 8 seconds. Running Recast at 768p on the full 20 seconds costs $7.50. Trimming to 8 seconds first costs $3.00, then you rejoin the spans in a timeline render. The trim is $0.02 per job and the timeline render is billed at its own compute, so the trimmed route saves roughly $4.40 on a single clip before those small fees, and the saving scales with every revision.

Trim is a fixed-price step at $0.02 per job, so cutting the source before the edit is almost always cheaper than editing the whole file. Check the trim and timeline rates in the docs before you quote a final figure to a client.

Limits

Both models round a fractional input up to the next whole second, so a 10.1-second source bills as 11. The tables show the model amount. A workspace Agent Fee line can be added on top, so treat the job's own usage cost as the truth and check the live catalog before you budget; limits and rows change.