Genjutsu AI for a Hotel Lobby-style video
Start with the result you want: carry a performance from an existing clip into a new scene, or create a fresh duo performance from two portraits. This guide explains where Genjutsu fits and how the two-photo creator on this site differs.
What is Genjutsu AI?
Genjutsu is a Higgsfield tool for transforming existing video. Its Motion Transfer mode uses a source performance to guide movement, camera motion and timing in a rebuilt scene. Object Swap targets selected elements, such as a character or outfit, while aiming to retain the surrounding shot. See the official Genjutsu page for the current controls and credit cost.
This site is an independent two-photo video creator. It is not Higgsfield or an official Genjutsu interface, and the creator below does not submit tasks to Genjutsu.
Prepare your reference material
For a reference-driven duo edit, start with a clip where both performers are easy to distinguish. Check the whole shot: crossed faces, extended occlusion and fast cuts make it harder to judge whether identities remain consistent.
Choose one clear portrait for each person and keep the references organized. Use a clip and images you have permission to use. Decide whether your priority is the existing performance, the setting, or a new scene before choosing a workflow. The template guide explains these different starting points.
Plan a two-performer edit
- Open Genjutsu on Higgsfield and add your source clip and image references.
- Choose Motion Transfer for a rebuilt scene guided by the original movement, or Object Swap for a targeted change.
- Make the intended person-to-reference mapping clear. If you use a prompt, describe each performer unambiguously and use the reference labels shown in that interface.
- Check the displayed generation cost and input limits before submitting. These can change; use the live interface rather than assuming a fixed allowance.
- Review the entire result, including identity consistency, hands, overlaps and sound. Save a useful version before experimenting again.
A planning note might read: “Use the first portrait for the performer who begins on the left and the second portrait for the performer who begins on the right.” Adapt that note if the performers exchange positions. It is an instruction to review, not a guarantee that every frame will preserve the mapping.
Create a new duo performance from two photos
If you want two people in an orange studio scene without preparing a source clip, open this site's two-photo creator. Add one adult portrait per slot, check the previews, sign in and review the credit requirement before generating.
The current preset requests an 8-second, 720p, 16:9 duo video. It generates a new performance rather than copying the movement of your reference video. There is no source-video upload, custom prompt field or exact lyric control in this workflow.
Generated audio may accompany the result but is not guaranteed. The original Hotel Lobby recording is not included. Read the song and audio guide before choosing a soundtrack, and the step-by-step tutorial for task tracking and downloads.
Common questions
Can I use Genjutsu directly on this page?
No. Follow the official Higgsfield link for Genjutsu. The create button on this site opens our separate two-photo workflow; accounts and credits are separate.
Is Genjutsu free?
Higgsfield describes generation as credit-based and displays the cost before a run. Check its current offer in your account. This site's local photo preview does not generate a video; video generation uses the credits displayed in our creator. See the free-preview guide.
Will the two-photo creator reproduce the exact dance?
It creates new movement from the prepared scene. If following a specific source performance is essential, evaluate a reference-video workflow and check its result against your source.
What should I check before sharing?
Watch both faces throughout the saved clip, listen to the complete audio and preview any vertical crop. Ask the participants to review the result and make clear that the performance is AI-generated. For help with a task created here, use support.