What Is Genjutsu AI?

A reference video can be more than something to watch: it can be the starting point for a different scene. Here is where Genjutsu AI fits, what it can change, and what still needs your judgment.

Try Genjutsu AI on the homepage — pick a viral template or upload your clip, add reference photos, and check the estimated credits before generating.

What is Genjutsu AI?

Genjutsu AI is an independent online AI video platform at genjutsu.im that uses third-party AI models. It runs in your browser and turns source footage and reference images into inputs for a new video, rather than asking you to build every change on an editing timeline.

Its two core video workflows are Motion Transfer, for a scene guided by an existing performance, and Object Swap, for a replacement within an existing shot. Video templates provide ready-made starting points; image creation and editing tools are also available. These are generative tools, not a promise that every source detail will survive unchanged.

It suits creators who prefer to stay off camera, recurring-character or pet accounts, and sellers or editors exploring product-video ideas. Editing experience is not a prerequisite for a first generation, though reviewing and finishing the result still matters. Access does not require a subscription: one-time credit packs are available.

What can it do? Two reference-led video workflows

Choose Motion Transfer when the movement is what you want to keep. Choose Object Swap when a particular element is what you want to change.

• Motion Transfer on the homepage: give a different character, pet, or scene the movement and timing of your reference. A short, readable gesture is a better first test than a crowded sequence with frequent occlusion.
• Object Swap on the homepage: use your existing shot as the starting point and provide references for a replacement person, product, outfit, prop, or background. Check both the replacement and the parts you meant to preserve.

Who is it for? Use cases and what to check

The useful distinction is not just who appears in the video, but what makes the result usable for your project.

For background comparisons, do not feed one generated version into the next. Where exact labels or unchanged pixels are essential, keep original artwork, real footage or conventional compositing in the finishing workflow.

How to start with one short scene

1. Choose one short clip of your own or a video template, plus clear reference images that show the intended subject. Keep the original and follow the selected workflow’s upload requirements.
2. Decide on one change, choose the mode and tier, and check the credit estimate. Prompts are optional: leave the toggle off or add a short sentence for a specific change. Lite accepts 4–15-second clips; Standard and Pro accept 4–30-second clips. Standard and Pro follow the source duration and framing, rather than offering arbitrary output length or aspect ratio.
3. Review the whole result, including hands, edges, text and sound. If another test is worthwhile, change one input and keep a note of it.

If a job stays queued or processing, check its history before paying for duplicates. Save the task ID, settings and a screenshot; use the support contact if the status remains unclear. A long wait alone does not explain the cause or billing outcome.

Credits, quality tiers, and free trial credits

Quality tiers select how a generation runs; they are not the names of paid plans. Lite and Standard produce 480p output, while Pro produces 720p output.

For a reference-led video of about 7 seconds, using a reference and output of similar length, the current estimates are:

These are examples, not a flat price for every video. Duration, workflow, and billing measurements matter. Check the estimate before each generation; the final charge follows actual billing. Standard and Pro may retain audio from the reference or include sound, so listen to the downloaded result before sharing it.

For a useful budget, count all attempts—not just the final keeper. Divide the credits spent across your tests by the number of clips you would actually use, and track review and editing time separately. A less expensive test can help you decide whether the scene is worth refining before choosing a higher-resolution run.

Eligible new users can receive 5 trial credits after email verification, valid for 1 day. The grant is subject to eligibility and issuance rules. Those 5 credits are not enough for the 7-second video examples above, including the 26-credit Lite example. A trial allowance is not a promise of a free full video.

No subscription is required: one-time credit packs are available, with optional monthly and yearly plans. Paid options use the names Starter, Basic, and Pro. These plans and credit packs are separate from the Lite, Standard, and Pro generation tiers. Check the current pricing and purchase options for allowances, validity, and billing terms instead of assuming that a plan name determines every generation setting.

Frequently asked questions about reviewing results

What should I change first if the subject looks right but the motion feels wrong?

Keep the subject references fixed and revisit the source action first. Compare the result with the original at the turn, gesture, or contact point that matters. If that moment is obscured or cut off in the source, test a clearer section before changing both the photos and prompt. Change one input per attempt so you can tell what helped.

How do I budget for tests rather than just one finished clip?

Set a total credit limit and reserve part of it for a keeper before starting. Record each attempt’s displayed estimate, actual charge, changed input, and whether the clip is usable. A new attempt needs its own budget; do not multiply a single 7-second quote into a guaranteed number of publish-ready videos. Stop when another retry would consume the credits you reserved for the final version.

How can I compare three background versions fairly?

Start each version from the same original clip and product references, and compare the same moments. Use a simple checklist: product silhouette, label, hand contact, camera position, and reflections. Evaluate the intended background change separately from accidental product changes. Do not use the first generated version as the source for the next comparison.

When should I stop generating and finish the clip in an editor?

Stop when the intended visual change works and the remaining corrections are specific editing tasks. A precise cut, a caption, an audio adjustment, or a label overlay may be better handled in your editing workflow than by another whole-scene generation. If identity or product shape is still wrong, do not treat an attractive first frame as an approved result.

Explore Genjutsu AI with a clear idea

Choose the part of the reference you love: the movement, the composition, or the subject you want to replace. Set a test budget and decide what a successful result would look like. Then open Genjutsu AI, review your inputs and estimated credits, and try one focused scene before committing to a larger project.

Make a Genjutsu AI video on the homepage — no software to install. Preview the result in your browser and download a watermark-free 480p or 720p MP4. Use only footage, images, likenesses, and audio you have permission to use.