Motion and timing stay intact
The source steers body motion, object trajectories, cuts and camera rhythm, so the restyle follows a performance that has already been signed off.
Drop in a source video and write the look you want instead. MuseGen restyles the footage, letting its movement, timing and camera path steer the generated version.
Generating needs an account and spends credits based on the model and length you pick. Each model sets its own source-clip limits, and failed tasks are refunded on their own.
Freeze each finished look on the left, then play that result on the right. Copy the direction, or send it to the restyle generator at the top of the page.
Look frameSingle continuous uninterrupted cinematic shot, no cuts, one flowing 15-second take. Shot on ARRI Alexa with anamorphic lenses, photorealistic live-action film, NOT animated, NOT a video game, NOT CGI render. On a private jet runway tarmac in bright daytime, a sleek private jet parked with its boarding stairs lowered. Cinematic color grading, crisp natural daylight, heat shimmer off the tarmac, clean reflective surfaces, filmic contrast, natural film grain, shallow depth of field. Photorealistic textures, real human skin, realistic materials and reflections.The single camera begins up close beside the boarding stairs where a few beautiful female flight attendants wait — elegant, professional, well-groomed women in crisp tailored white uniforms with welcoming smiles. There is no one else around, only the attendants. The man [imageReference] rides into frame on a premium futuristic motorcycle and comes to a normal, smooth, calm stop right in front of the camera — an ordinary gentle braking with no dust, no skid, no dramatic impact, just easing to a stop. He wears relaxed but refined vacation clothing — a clean fully-buttoned short-sleeve shirt, smart casual trousers, sunglasses, and clean white sneakers, looking effortlessly cool. The motorcycle is an elegant high-end design finished in deep glossy black and metallic red, premium and sophisticated.Without any cut, the same camera flows with him as he swings off the bike confidently and the female attendants greet him warmly. The camera continues unbroken, keeping him in a wider full view from a distance, as he steps beside the motorcycle and presses a button on the bike — seen naturally from afar in the same wide framing, no close-up. The button lights up and the motorcycle responds: its panels smoothly fold and retract, the frame compacting and reconfiguring with precise futuristic mechanical motion, wheels tucking in, the whole machine collapsing down and reshaping itself into a sleek premium rolling suitcase in the same matching deep black and metallic red finish, a telescopic handle rising up. The same camera orbits gently around the transformation, never cutting. The female attendants react with amazement.Still in the same continuous unbroken take, the camera follows as the man casually grabs the suitcase handle, gives it a confident look, and walks at a relaxed unhurried pace pulling the rolling suitcase behind him to the boarding stairs. The shot ends just as he steps onto the first steps and begins to ascend, the camera trailing smoothly behind and to the side — all in one single flowing continuous shot with no cuts anywhere.Photorealistic, premium, clean luxury commercial film look, one smooth steady continuous camera move, restrained cinematic color palette, filmic grain.
Let a performance or camera move you already have carry the motion, then rewrite the art direction, the location, the materials, the lighting or how the subject is rendered.
Right now Seedance 2.5, Seedance 2.0 and Gemini Omni Flash take a source video for restyling. They vary in length limits, reference caps, how they treat audio and how tightly they stick to the original structure.
Four reasons footage beats rebuilding a performance from a text prompt.
The source steers body motion, object trajectories, cuts and camera rhythm, so the restyle follows a performance that has already been signed off.
Go from live action to illustration, cartoon, film stock, fantasy or a material-led aesthetic with no frame-by-frame repainting.
Your clip provides the framing and the spatial layout, which gives the model far more to lean on than a prompt when several things move together.
Spin a single good performance into campaign, social or concept variants while its key motion and edit timing stay the same.
Supply the motion, describe the new world, then check the result against the original.
Pick footage whose movement, timing or camera path you want to keep. Short, steady shots give the model a cleaner structure to track.
Name the style, location, materials, light and subjects that should change, then spell out which motion, framing and beats have to carry over.
Generate a short restyle, compare its movement to the original, and firm up the prompt wherever something drifted before you render again.
Six ways video to video turns good motion into a brand-new creative asset.
Turn filmed acting into anime, painted 2D, stop-motion or stylised 3D while the performer's overall motion and the shot's pacing survive.
Recycle a product spin, a hand demo or a camera sweep and change the set, the light, the season and everything around the item.
Keep the choreography and edit rhythm of the original take, then try surreal, graphic, vintage or cinematic treatments around it.
Build several visual concepts from one short clip for different posts, launches or audiences, and never ask the talent to do the action twice.
Turn rough blocking footage into a quick concept for a creature, a set, a costume or an effects shot before you lock the real production.
Move an evergreen clip into a winter, summer, launch or event look while the motion that already performs stays put.
Style-led video prompts from the MuseGen gallery. Choose a result to load its look into the generator, then adapt it to your own source video.
Video to video runs on the models that accept a source video. Credits follow the chosen model and length, and failed generations are refunded automatically.
For first-time AI creators
$19.9
$179 billed yearly
Save $60 compared to monthly
12,000 credits granted for the full year
Estimated monthly output
What you get
For everyday AI creation
$49.9
$419 billed yearly
Save $180 compared to monthly
30,000 credits granted for the full year
Estimated monthly output
What you get
For ambitious AI projects
$99.9
$719 billed yearly
Save $480 compared to monthly
60,000 credits granted for the full year
Estimated monthly output
What you get
A video to video AI generator makes a fresh clip from footage you supply plus written direction. Rather than dreaming up all the motion from text, it takes the source's movement, timing, layout or camera path as the guide for the transformation.
Image to video begins with one frozen frame and must imagine everything that follows. Video to video starts from motion that is already there, which is the better choice when a performance, a camera move, an object's path or the edit rhythm should stay while the look changes.
Today the generator offers video to video through Seedance 2.5, Seedance 2.0 and Gemini Omni Flash. In Restyle mode the model picker hides any model that cannot take a source video.
Choose a short, clean clip where the movement is easy to read and compression is light. Steady framing and a subject nothing passes in front of are simplest to preserve. Rapid cuts, heavy blur and tiny subjects leave the model with weaker structure.
It varies by model. Seedance 2.5 currently takes source clips of 2 to 30 seconds, Seedance 2.0 takes about 1.8 to 15.2 seconds, and Gemini Omni Flash takes a single source clip of up to 24 seconds.
The source acts as a guide, not a frame-accurate filter. Overall movement, timing and layout tend to transfer well, while hands, faces, small props and very fast action can shift. List whatever has to stay unchanged in the prompt.
You can tell the model to keep the subject and motion and swap the surroundings, though a generative restyle may still nudge lighting, outlines or fine details on the subject. A clearly visible subject and an explicit keep-this instruction help it separate the two.
Yes — style conversion is one of the most common reasons to use video to video. Describe the rendering style, linework, palette, lighting and materials, and directly ask the model to hold on to the original performance and camera timing.
That depends on the model and what you ask for. Seedance can create native sound for the new clip, but keep your own copy of the original track whenever its exact dialogue, music or timing has to survive into the final edit.
The visual change you asked for may be fighting the motion constraint, or the clip may be full of fast cuts and objects blocking the subject. Trim the clip, simplify the change, and list the motion and layout that must stay before adding smaller style notes.
Uploading a source and setting up the restyle costs nothing. Rendering needs an account and credits; new accounts get 15 free credits, and any task that fails is refunded automatically.
Paid plans include commercial use under the current terms. You still need to own or license the source footage, people, music, logos and anything else you upload; an AI restyle does not stand in for those permissions.
Upload a clip whose movement you like, describe the world you would rather see, and render a new version without reshooting the performance.
Restyle a Clip