Restyle
Take a plain clip and give it a look — film stock, anime, claymation, a specific decade, a particular palette. The action stays recognisable while the visual language changes completely.
Video to Video
AI video restyling
Upload existing footage and describe a new visual style — Seedance 2.0 uses your clip as reference to generate a restyled AI video. Transform the look without re-shooting.
MP4 or MOV, 2–15s each, 15s total, up to 200MB. Requires R2 with a public domain — the model fetches the file over the network.
Model
Input modeReferences
Reference videos 0/3
MP4 or MOV, 2–15s each, 15s total, up to 200MB. Requires R2 with a public domain — the model fetches the file over the network.
0/2000
4s15s
Outputs
Audio
With audio
Synchronized sound, voice and music
Credits required:36
Your generated video appears here. Describe a shot and hit Generate.
My creationsDownload
Generation tips
- - Upload existing footage and describe a new visual style — Seedance 2.0 uses your clip as reference to generate a restyled AI video. Transform the look without re-shooting.
- - MP4 or MOV, 2–15s each, 15s total, up to 200MB. Requires R2 with a public domain — the model fetches the file over the network.
- - Generation time varies with the model, clip settings, and provider load.
- - Name the subject, the action, and the camera move. Vague prompts give vague shots.
- - Longer clips cost more credits — start at 5s to test a look, then extend.
- - Audio is generated with the picture; describe the sound you want in the prompt.
Video to video — keep the motion, change everything else
Video to video takes footage you already have and rebuilds it: a new grade, a new setting, a new style, a different era — while the original motion survives. The performance, the timing and the camera move are yours; what surrounds them is regenerated. It is the closest thing to re-shooting a scene without a camera.
This is a genuinely different operation from filtering. A colour LUT changes pixels; Seedance 2.0 re-generates the scene while conditioned on the source clip’s motion, so it can put your subject somewhere the camera never went and light it accordingly. On FrameAI you can feed in up to three reference clips totalling fifteen seconds, combine them with reference images and a text prompt in the same request, and export up to 4K with synchronized audio. Commercial use of paid-plan output is governed by the Terms of Service.
- Original motion and timing preserved
- Restyle, relocate or re-light a scene
- Combine with images and text in one request
Three jobs video to video does well
All three share a premise: the movement in your source is worth keeping, and everything around it is negotiable.
Relocate
Keep the performance and put it somewhere else. A walk filmed in an office becomes a walk through a rain-soaked street, with the lighting on the subject rebuilt to match the new environment.
Borrow the move
Use a clip purely as a motion reference — its camera behaviour and pacing applied to an entirely different subject. The source becomes a direction rather than a picture.
What survives the transformation, and what does not
What reliably carries through is structure: the timing of an action, the trajectory of the camera, the rough spatial layout, the rhythm of the cut. That is why a restyled clip still feels like your clip — the choreography is intact. What does not carry through unchanged is fine detail. Small text, intricate logos, precise facial micro-expression and delicate hand articulation are all regenerated, and regeneration means variation. If a detail is legally or commercially load-bearing, pin it with a reference image rather than trusting it to survive the round trip.
How far you push the prompt determines how much is preserved. A modest request — regrade this to look like evening, keep everything else — stays extremely close to the source. An aggressive one — turn this office into a submarine — necessarily rebuilds most of the frame, and structural fidelity loosens as it does. The useful habit is to state explicitly what must not change, then ask for the transformation, and to work with source clips that are well-exposed and reasonably stable, because the model inherits your source’s problems along with its motion.
Reliable to preserve
Action timing, camera trajectory, spatial layout, overall pacing, the general shape and position of the subject.
Pin these explicitly
Readable text, exact logos, precise brand colour and facial identity. Attach a reference image for anything that must be exactly right.
Model specifications
What Seedance 2.0 actually delivers on FrameAI
Published limits, not marketing rounding. Every number below is enforced by the generator before your credits are reserved, so a job that would exceed a limit is rejected instead of failing halfway.
- Up to 4K
- Output resolutions — 480p / 720p / 1080p / 4K
- 4–15 sec
- Clip length per generation, at 24 fps
- Native audio
- Sound effects and ambience generated in sync
- 9 + 3
- Reference images plus reference videos, one request
- 7 ratios
- Supported aspect ratios — Auto, 16:9, 4:3, 1:1, 3:4, 9:16, 21:9
- Commercial use on paid plans
- Subject to the Terms of Service
Why creators run Seedance 2.0 on FrameAI
The model is ByteDance’s. What FrameAI adds is the part that decides whether it is usable at work: honest limits, honest billing, and output you are allowed to ship.
Every input in a single pass
Text, up to nine reference images and up to three reference video clips are fused in one generation, not stitched afterwards. There is no separate text-to-video mode to switch into — you add whatever material you have and the model reconciles it.
Sound generated with the picture
Seedance 2.0 writes footsteps, room tone, impacts and score alongside the frames, locked to the action. Most models hand you a silent clip and leave the sound design to you; here the export is already finished.
Seven ratios, up to 4K
One prompt covers a 9:16 vertical cut for Reels, a 16:9 master for YouTube and a 21:9 anamorphic frame for a title sequence. Resolution runs from 480p up to 4K, so the same take can serve a cinema-shaped frame or a phone.
Credits you can predict
The exact credit cost is calculated from your duration, resolution and ratio and shown before you press generate — never after. If a generation fails on the provider side, the reserved credits are returned automatically.
Built for iteration, not for queuing
Seedance 2.0 Fast and Seedance 2.0 Mini exist for the twenty drafts nobody sees. Rough a shot at 480p for a fraction of the credits, then re-run the take that worked at full resolution on the flagship model.
Yours to sell
Paid-plan output may be used commercially under the Terms of Service. Free-credit output is for evaluation only.
Transform a clip in three steps
The source does the choreography, so your prompt only has to describe the destination.
- 1
Upload your source clip
Up to three clips, fifteen seconds combined. Trim to the segment that matters — a shorter, cleaner source produces a better result than a long one with dead air.
- 2
Describe the destination
Say what the result should look like and, just as importantly, what must stay the same. Add reference images for any element that has to be exact.
- 3
Generate and compare
Review against the source. If it drifted too far, soften the prompt; if it barely changed, push harder. One or two passes usually finds the balance.
Where video to video earns its keep
The common thread: existing footage that is nearly right and would be expensive to shoot again.
Rescuing and repurposing an archive
Old campaign footage gets a contemporary grade and a new setting instead of a reshoot. A library that felt dated becomes usable again.
Style variants for testing
One filmed performance, five visual treatments. Testing which look performs no longer means paying for five separate productions.
Turning a rough take into a finished one
A phone-shot reference of the action you want becomes the motion spine for a polished, generated version — the fastest way to direct a shot precisely.
Video to video — frequently asked questions
What is video to video?
It regenerates an existing clip according to a text prompt while preserving the source’s motion and timing. You can restyle it, relocate the scene, change the lighting or the era — the choreography stays, the rest is rebuilt.
How long can my source clip be?
Up to three reference videos with a combined length of fifteen seconds. Trim to the section that matters — a tight source consistently outperforms a long one.
Is this just a filter?
No. A filter maps pixels to new pixels. This regenerates the scene conditioned on your clip’s motion, which is why it can change the environment, the lighting direction and the materials rather than just the colours.
Will faces and logos stay accurate?
Not automatically — fine detail is regenerated, and regeneration varies. Attach a reference image of any face, logo or product that has to be exactly right, and keep the prompt’s ambition proportional to how much must be preserved.
Can I combine a source clip with reference images?
Yes, in the same request — up to nine images alongside your clips. This is the recommended approach when motion comes from the video but identity or branding must come from a still.
What quality of source footage do I need?
Well-exposed and reasonably stable. The model inherits your source’s problems along with its motion, so severe blur, heavy compression or violent shake will show up in the result.
Keep going with the rest of the toolkit
Each tool is the same Seedance 2.0 model pointed at a different job. Move between them freely — your credits, history and exports are shared.
Generate your first Seedance 2.0 video now
Write one line, pick a ratio, press generate. No editing suite, no render farm, no post-production pass — a finished clip with sound, ready to download.
- Free credits on sign-up
- Paid commercial use
- Refunded if a generation fails