How to use Genjutsu Video in five steps
1. Sign in with Google or an email link, and buy a credit pack on the pricing page if your balance is empty. 2. Open the studio and choose a workflow template: Bring an image to life, Restyle a video, Swap an object, Animate → restyle, or Animate → pass it on. 3. Upload your MP4, plus a character image for motion transfer or an optional object image for object swap. 4. Pick a model, choose a preset or write a description, and check the credit quote. 5. Generate, then download the MP4 from My results.
Prepare your source video
Use a single continuous shot of 3–10 seconds. Keep one person visible, with clear head and body proportions. Export a standard MP4 at 24–60 fps with both dimensions between 720 and 2160 pixels, and keep the file under 30 MB. Trim away cuts, title cards and long still frames before you upload.
Choose motion transfer, restyle or object swap
Motion transfer animates the character in your image using the performance in your video; the image also guides the background and overall look. Restyle keeps your whole clip, people and places included, and redraws it in a new style, so no character image is needed. Read the motion transfer and AI video restyle pages for model details. Object swap changes one object, product, outfit, background or the weather and keeps everything else; see the AI object swap page.
Give the model a clear character image
Use a sharp JPG, PNG or WebP image under 10 MB, or a JPG or PNG under 4.7 MB for DreamActor. Match a full-body clip with a full-body image. Avoid faces hidden by hands, busy crowds and extreme angles.
Write a focused description
For motion transfer, describe the scene the character should perform in: “The character performs the same gestures as the reference video in a softly lit studio.” For restyle, describe the look: “Restyle the whole video as a loose watercolor painting. Keep the motion and camera.” Describe styles, places and light rather than the people in your source clip; requests to swap faces or imitate a real person are refused.
Review, download and continue
Successful videos appear automatically in My results, where you can download them, delete them, or use one as the source for the next step of a workflow.
Troubleshooting
If an upload is rejected, check the length, the file size and the pixel dimensions first. If a model rejects your inputs or a job fails, the held credits return automatically; try a clearer image, a steadier clip or a different model. Write to support@genjutsuvideo.com with the job ID if anything looks wrong.
Workflow templates explained
Bring an image to life is a single motion transfer step. Restyle a video is a single restyle step with no character image. Swap an object is a single object swap step. Animate → restyle runs motion transfer first and then redraws that take in a new style. Animate → pass it on animates one character and then uses that take as the performance for a second character image. Every step shows its own quote, and the next step only starts when you have reviewed the previous result and choose to continue.
Choosing a model when you use Genjutsu Video
Start motion transfer with Kling 2.6, the default, and move to Kling 3.0 when the take matters. Use DreamActor v2 when facial expression is important and your image is a JPG or PNG under 4.7 MB. Use One-to-All for quick, low-cost drafts at 480p. For restyle, Lucy Restyle redraws the whole clip at 720p. The model picker shows the credits for five seconds next to each name.
Descriptions that work
Good descriptions are short and specific about the scene or style: “The character performs the same movement on a concert stage with spotlights and light haze.” “Restyle the whole video as stop-motion claymation with handmade sets. Keep the motion and camera.” Avoid describing the person in your source clip, avoid brand and studio names, and keep the instruction to preserve motion and camera when restyling.