Genjutsu lists at $0.154 per second on the Higgsfield API, and Genjutsu Light lists at $0.074, which puts the light tier at 48% of the standard rate. A 15 second motion transfer is $2.31 on standard or $1.11 on Light. A full 30 second pass, the ceiling standard Genjutsu accepts, is $4.62.
Everything about the pricing follows from one fact: Genjutsu does not generate a performance, it transfers one. The length of the video you feed it is the length you pay for.
If you are signing up to run this against a product or a character, the referral route takes up to 50% off for three hours. The link is here, and we may earn a commission if you subscribe through it, at no extra cost to you.
The Two Rates
Higgsfield publishes per second list prices on the API price sheet, and Genjutsu is the most expensive line on it. Both tiers run 2 modes at up to 1080p.
| Model | List price per second | 4 seconds | 15 seconds | 30 seconds |
|---|---|---|---|---|
| Genjutsu Light | $0.074 | $0.30 | $1.11 | $2.22 |
| Genjutsu | $0.154 | $0.62 | $2.31 | $4.62 |
That is the number without a subscription attached. Inside the plan, Genjutsu runs on credits like every other model, and the credit cost for the specific clip is shown on the Generate button before it renders. I am not going to publish a credit rate for it that I have not read off Higgsfield's own comparison table, because a made up rate is worse than no rate.
You Are Paying For the Reference Video, Not the Output
Here is the part people get wrong when they budget a test. Genjutsu's inputs are a reference video plus reference images, and the output inherits the length of that reference. Feed it a 30 second clip and you are billed 30 seconds. Feed it a 15 second trim and you are billed 15.
So splitting a long transfer into two shorter jobs does not save a cent. Thirty seconds is thirty seconds either way. What saves money is trimming the reference to the frames that actually matter, before the first render rather than after the tenth.
That leads to the workflow the price sheet is begging for:
- Trim the motion reference to the 15 seconds you need. This alone halves a standard job.
- Test the transfer on Genjutsu Light, at $0.074 a second, until the tracking holds.
- Run the approved take on standard Genjutsu, which is the only tier with Restyle and the longer reference ceiling.
Twenty Variants of a 15 Second Transfer
Agencies do not run one of these. They run a batch and keep the two that work. Priced on the API list rates:
| Batch | Genjutsu Light | Genjutsu |
|---|---|---|
| 5 variants, 15s | $5.55 | $11.55 |
| 10 variants, 15s | $11.10 | $23.10 |
| 20 variants, 15s | $22.20 | $46.20 |
An hour of finished 30 second transfers, 120 of them, is $266 on Light or $554 on standard. At that volume the gap between the tiers is a month of a mid tier subscription, which is the argument for testing cheap and finishing expensive instead of picking one tier and living with it.
Lite Against Standard, Mode by Mode
The changelog entry from October 19, 2026 explains what Light actually is. Motion transfer and object swap work the same way as standard. Light accepts a reference video from 4 to 15 seconds and up to 10 reference images, at 720p or 1080p. For longer videos, more references, or Restyle, Higgsfield tells you to switch to standard.
Restyle is the one real functional wall. It runs in Motion transfer mode with reference videos from 4 to 30 seconds, and Light cannot do it at any price. If Restyle is the whole reason you are on the tool, the $0.074 rate is irrelevant to you and the 30 second ceiling sets the budget.
How It Compares With a Generative Model
Kling 3.0 lists on the same API at $0.107 per second for the Standard mode, so the same 15 second shot is $1.61 there against $2.31 on Genjutsu standard. Genjutsu costs about 1.4x as much per second.
What you buy for that 44% premium is skipped retakes. A generative model has to invent the performance, the camera move and the product placement, and you burn generations until all three land. A motion transfer locks the performance and the camera to your reference on the first pass. If your prompt hit rate on a generative model is one in four, Genjutsu at 1.4x is the cheaper route to a usable clip, and that math is the entire case for the tool.
Where it loses: anything the reference does not contain. New camera angles, new blocking and new characters all mean a generative model, and the model guide covers which ones to reach for.
The verdict: budget Genjutsu by reference length, not by clip count, and test on Light at $0.074 a second before paying $0.154. The two tiers differ by mode and reference ceiling, not by output quality. For the preset side of the tool, the Genjutsu presets guide covers what the 30 plus presets actually do.