Video to Video, Free: Restyle, Edit and Extend Any Clip, No Watermark
Upload a clip and say what should change. The motion, the camera and the timing stay; the look, an object, a person or the ending is yours to rewrite. New accounts start with free credits. Free accounts download without a watermark.
Perfume swapped for face creamSeedance 2.0Grey previz rendered as a car filmSeedance 2.5 · 16:9 · 15sBlockout turned into a desert chaseSeedance 2.5 · 4:3 · 15sTen seconds added to a climbSeedance 2.5 · 10s
Restyle a video into anime, 3D or a new look and keep the motion
Yes. Attach the clip and describe the look you want; the model reads the movement, the camera path and the timing from your footage and redraws the surfaces, the light and the style on top. It works in both directions: live action into anime or claymation, and a rough grey 3D blockout into finished photoreal footage, which is what the demo beside this text does. Free credits run it on Seedance 2.0 Mini at up to 720p; Seedance 2.5 holds a longer clip together, up to 30 seconds. Describe the style rather than naming a studio, a franchise or a character: copyright and trademark was the reason given for 72 of the 176 refusals counted on this site between October 10 and November 9, 2026.
Steps, prompt and settings
How to do it
Attach the source clip; keep it under 15 seconds on Seedance 2.0 and under 30 on Seedance 2.5.
Say what stays: the motion, the framing, the camera path and the timing.
Describe the new look with materials, light and palette, not with a studio or franchise name.
Generate at the length of the source; a restyle that runs longer than its clip has to invent the rest.
Prompt
Keep the motion, framing, camera path and timing of the source video exactly. Restyle it as hand-painted 2D animation: flat cel shading, clean ink outlines, a warm pastel palette, soft evening light. Do not add or remove anything in the scene.
Settings
Model: Seedance 2.0 Mini on free credits (up to 720p) · Seedance 2.5 for clips up to 30 seconds
Inputs: one source clip, optionally one still of the target style
Duration: match the source · Audio: on or off
This recipe is for your own run. The demo beside it was made with the prompt printed under it.
Input
▶ 1
Seedance 2.5
Prompt: Render the white model: upgrade the car body to a deep blue gradient metallic paint, with dark areas in navy blue and deep sapphire blue, naturally transitioning on lit surfaces to cobalt blue, bright blue, and electric blue, with fine metallic particles, clear-coat reflections, and flowing highlights. Upgrade the mechanical structure to cool gray metal, brushed metal, matte alloy, and sprayed steel materials. Upgrade the interior and seats to dark brown/walnut brown premium leather, with fine texture, slight wrinkles, and soft reflections. The glass and lamp covers should have a realistic transparent refractive texture … (Seedance 2.5 · 16:9 · 15s · 720p)
Swap an object in a video, such as a product, a car or a prop
Yes, and it is the job this entry does best, because what you keep is the shot and what you change is small. Give the model one still of the replacement, name the object to remove and the image it comes from, and leave the rest of the sentence empty. The pair beside this text was made exactly this way on Seedance 2.0: a 5.5-second clip of a gift box with a perfume bottle in it, one still of a face cream, and the one-sentence instruction printed under the demo. The camera, the set and the lighting stayed because nobody asked them to move.
Steps, prompt and settings
How to do it
Attach the clip that has the object in it.
Add a still of the replacement, shot as cleanly as possible: one object, plain background.
Name the object to replace and the image that replaces it; do not describe the scene.
Run it; if the object drifts, shorten the source clip or crop the still tighter.
Prompt
Replace the glass on the table with the can from the image. Keep the hand, the table, the lighting and the camera movement unchanged.
Settings
Model: Seedance 2.0 or 2.0 Mini on free credits · Seedance 2.5 for a localized edit
Inputs: one source clip plus one reference still
Real pair: the first slide of the examples at the top of the page
This recipe is for your own run. The demo beside it was made with the prompt printed under it.
Input
▶ 1
1
Seedance 2.0
Prompt: Change the perfume in the video gift box to the face cream from the image. (Seedance 2.0 · 16:9 · 4s · 720p)
Use a clip as the motion and camera guide for a new video
Yes. Here the clip is not edited at all: it is a guide. The model copies its camera move, pacing and blocking, and the prompt and the reference stills decide what is actually in the picture. That is how a previz, a 3D blockout or a phone rehearsal becomes a finished shot, and it is the way the demo beside this text was made: one untextured chase animation as the guide, plus stills that fix the look of the motorcycle and the SUV. Seedance 2.5 takes up to ten clips and 30 images in one job; Seedance 2.0, Fast and Mini take up to three clips and nine images. To move one performance onto a single character still, Kling Motion Control is the specialist.
Steps, prompt and settings
How to do it
Attach the clip that has the movement you want, then the stills of the subjects.
Name each file by its number: “@Video 1” for the guide, “@Image 1” and “@Image 2” for the looks.
Say that the clip controls camera movement, pacing and blocking, and that the images control appearance.
Set the output to the length of the guide clip so no extra movement has to be invented.
Prompt
Use @Video 1 as the only guide for camera movement, pacing, framing and blocking. Turn it into a realistic night street scene: the runner is the woman in @Image 1, wearing the jacket from @Image 2. Wet asphalt, neon reflections, handheld feel. Do not copy the look of @Video 1, only its motion.
Settings
Model: Seedance 2.5 (up to 10 clips and 30 images) · Seedance 2.0 Mini on free credits (up to 3 clips and 9 images)
Inputs: one guide clip plus one or more reference stills
Duration: the length of the guide clip · Resolution: 720p on free credits and Lite, 1080p on Pro and Premium
This recipe is for your own run. The demo beside it was made with the prompt printed under it.
Input
▶ 1
Seedance 2.5
Prompt: Use @Video 1 as the white-model control input for the entire film, transforming the white-model previz into a refined, realistic chase scene with a strong commercial blockbuster feel, ultimately generating a 15-second highly realistic action film short. Use the reference images as subject appearance controls: @Image 1 controls the motorcycle’s model, color, structural proportions, and materials; @Image 2 controls the SUV’s model, color scheme, exterior silhouette, and details … (Seedance 2.5 · 4:3 · 15s · 720p)
Extend a video past its last frame
Yes, on Seedance 2.5, the one model here with an extend mode. Attach the clip, start the instruction with “Extend the video by 8 seconds” and describe what happens next; the model picks up on the last frame and returns 4 to 30 seconds of new footage with the same subject, set, light and sound. The demo beside this text is a paid user’s run from October 13, 2026: 10 seconds added to a 10-second clip at 720p, finished 5 minutes 25 seconds after it was submitted. The result is the new segment only, to place after the original in your editor. Extending needs a paid plan or a credit pack, because Seedance 2.5 is not on free credits.
Steps, prompt and settings
How to do it
Attach the clip to continue, 2 to 30 seconds long.
Begin the instruction with the word “extend” and the seconds to add.
Describe the next action, and say that camera, lighting and sound carry on from the last frame.
Place the returned segment after the original clip in your editor.
Prompt
Extend the video forward by 5 seconds. Continue directly from the last frame with the same camera, lighting and sound: the cyclist rides off the far end of the bridge into a narrow town street, no cuts.
Settings
Model: Seedance 2.5, paid plan or credit pack; Seedance 2.0 Mini and the other free-credit models have no extend mode
Inputs: the clip to continue, as the only source
Duration: the seconds to add, 4 to 30 · Billed on source plus added seconds
This recipe is for your own run. The demo beside it was made with the prompt printed under it.
Input
▶ 1
Seedance 2.5
Prompt: [Generation Goal]
Extend the source video forward by 10 seconds as a live-action cinematic climbing sequence. The central subject is Elion, and the primary event is Elion's exposed mechanical shoulder failing mid pull-up: the servos seize, that arm tears loose from the pipe, and he drops to hang by one flesh-and-blood arm over the open shaft.
[Source Video]
The source video defines Elion's identity, clothing, the mechanical shoulder, the rusted pipe, the wall plate, the lighting, and the film grade. Continue directly from the source video's last frame: close-up, Elion's chin just above the round rusted pipe, eyes looking up, jaw set, his left hand clamped over the top of the pipe beside the bolted steel wall plate, the polished steel mechanical shoulder showing through the torn dark jacket at the lower left of frame. Hold this composition, framing, and light for the opening moment before any new action begins.
[Characters]
Elion — tousled black hair, pale freckled skin, grey-blue eyes, torn charcoal canvas jacket. His right shoulder is mechanical: layered polished steel plates and exposed joint actuators with a thin orange wire, visible through the torn fabric. His right hand grips the pipe further to the left, just outside the frame. His left arm is flesh and blood. The output contains only one Elion throughout.
[Scene]
An abandoned brutalist concrete shaft: board-formed concrete walls, a square rusted steel plate bolted flat to the right-hand wall, and a single thick rusted pipe extending from the plate horizontally out into the open shaft. Below, the shaft falls away many storeys into deep shadow. Elion is a small figure against the enormous wall. The shaft is empty; no other people are present.
[Stage 1 | 0-3 seconds]
Initial state: the source video's last frame, exactly as described above.
Primary event: Elion clenches his teeth and strains to haul himself higher, neck tendons standing out. A rising servo whine comes from the mechanical shoulder, then a grinding stutter. A short burst of small orange sparks spits out through the torn jacket, and a thin wisp of smoke curls from the joint. The steel plates jerk twice and lock rigid.
End state: Elion's eyes snap wide toward his own shoulder, the shoulder frozen and smoking, chin still level with the pipe.
[Stage 2 | 3-5 seconds]
Continue from the previous stage: same close-up, same pipe and wall plate, left hand still clamped beside the plate.
Primary event: The mechanical arm goes dead and falls away from the pipe. Elion's whole body drops in one hard jolt, his face sliding down out of frame below the pipe. His left hand stays clamped on the pipe as his full weight snaps onto it: the arm whips straight, the pipe dips and shudders, rust dust bursts off its surface.
End state: Only Elion's left hand and straightened forearm remain in frame, locked on the pipe beside the wall plate.
[Stage 3 | 5-8 seconds]
Cut to a wide shot from out in the open shaft, level with the pipe, looking back at the right-hand wall.
Primary event: Elion hangs by his left arm alone from the pipe beside the wall plate, a small figure against the vast concrete wall, legs dangling over the dark drop. His body swings out away from the wall and back in a slow pendulum arc. The mechanical arm hangs limp at his side, twitching once, a last spark falling and fading into the dark below.
End state: The swing settles, Elion hanging one-armed, body turning slightly, the drop beneath him.
[Stage 4 | 8-10 seconds]
Cut to a close-up on Elion's left hand and forearm on the pipe, the wall plate at the right edge of frame.
Primary event: The fingers are white at the knuckles, the forearm tendons trembling hard, the whole arm shaking under his weight. The fingers slide a few millimetres along the rusted surface, then clench and hold.
End state: Elion's left hand still wrapped around the pipe, the arm visibly trembling, rust flakes drifting down past the frame.
[Visual Style]
Match the source video: 35mm film look, visible organic grain, natural lens softness toward the frame edges, neutral natural colour, grey concrete stays grey, rust stays brown. The sparks are the only warm, bright colour. Skin with real pores, sweat sheen, and grime. Motion blur on the drop. All action plays at real-time speed, with no slow motion.
[Audio]
Diegetic only: <Elion's strained breathing through clenched teeth>, <a rising mechanical servo whine>, <a harsh grinding stutter of gears>, <a sharp electrical crackle and spark hiss>, <a ragged cry as he drops>, <a hard metallic clang and groan as the pipe takes his full weight>, <the wall plate bolts creaking>, <rust flakes ticking down into the dark>, <fast, shaking breaths>, low hollow wind moving through the shaft. NO BGM.
[Maintain Consistency]
Keep Elion's identity, torn jacket, and mechanical right shoulder consistent with the source video. The mechanical arm stays limp after it fails. The left hand stays on the pipe beside the wall plate from Stage 2 to the end. The pipe extends from the wall plate on the right-hand wall, and the drop stays below throughout. (Seedance 2.5 · 10s · 720p)
Replace a person in a video with another character
Yes. Attach the clip, add one still of the character who should appear instead, and name the swap in a sentence. The model keeps the movement, timing, camera and lighting of the original and re-renders the person. On free credits this runs on Seedance 2.0 Mini at up to 720p; Seedance 2.5, on a paid plan or a credit pack, does it as a localized edit that leaves the rest of the frame untouched. One limit to know before you start: a source clip that shows real people can be refused by the model’s content check. It happened to 21 of the 1,109 video-to-video jobs run here between October 10 and November 9, 2026, so a stylized or clearly fictional replacement is the dependable choice, and a refused job is not charged.
Steps, prompt and settings
How to do it
Attach the source clip, 2 to 15 seconds long; the composer switches to video to video.
Add one clean still of the replacement character as a reference.
Write the swap as one instruction: who goes out, who comes in from the image, and that everything else stays.
Generate, compare against the original in your library, and reuse the same still for every further shot.
Prompt
Replace the man in the video with the character from the image. Keep his movements, timing, framing, camera motion and lighting exactly as they are; change nothing else in the scene.
Settings
Model: Seedance 2.0 Mini on free credits (720p) · Seedance 2.5 for a localized edit on a paid plan or credit pack
Inputs: one source clip plus one reference still of the character
Output length follows the model: up to 15 seconds on Seedance 2.0, up to 30 on Seedance 2.5
Prompt recipe: no clip attached, paste it and run it.
Change the background or setting behind the action
Yes, without a green screen. Describe the new place and say that the people, their movement and the camera stay. The model re-renders the environment around the action: a kitchen becomes a studio, a car park becomes a coastline, noon becomes night. On Seedance 2.5 this is a localized edit that keeps the subject stable; on the Seedance 2.0 family, including Mini on free credits, the scene is re-rendered from the clip, which suits a full restyle more than a one-object fix.
Steps, prompt and settings
How to do it
Attach the clip; a still of the new setting is optional but keeps the look consistent across shots.
State what stays: the people, the action, the framing and the camera path.
Describe the new setting and the light that comes with it in one or two clauses.
Generate; for a series, reuse the same setting still so every shot lands in the same place.
Prompt
Keep the two people, their movements and the camera path exactly as in the source. Change the setting to a rooftop at dusk with city lights behind them; match the new light on their faces and clothes.
Settings
Model: Seedance 2.5 for a localized background edit · Seedance 2.0 / Mini for a full restyle
Inputs: one source clip, optionally one still of the new setting
Resolution: 720p on free credits and Lite, 1080p on Pro and Premium for Seedance
Prompt recipe: no clip attached, paste it and run it.
Change clothes, hair or makeup and keep the person
Yes. This is the smallest kind of replacement: the person, the performance and the camera stay, and one attribute changes. Name the attribute and the new value, and say the face and the movement stay. A reference still of the garment or hairstyle helps the model match a specific look. The same caveat as for any clip with real people in it applies: the content check stopped 21 of 1,109 jobs here for that reason, and a refusal is not charged.
Steps, prompt and settings
How to do it
Attach the clip of the person.
Add a still of the garment, hairstyle or look if you need an exact match.
Write one change: “change only the jacket to…”, “change only the hairstyle to…”.
Generate; keep the source short so the change holds across every frame.
Prompt
Keep my face, expressions, movements and the room exactly as in the source. Change only my hairstyle to soft natural curls, shoulder length.
Settings
Model: Seedance 2.0 Mini on free credits · Seedance 2.5 for a localized edit
Inputs: one source clip, optionally one reference still
Output length follows the model, not the source
Prompt recipe: no clip attached, paste it and run it.
What an existing clip becomes
Video to video is the entry for footage that is almost right: the shot works, but its look, something inside it or its length has to change.
A finished look from rough footage: Push a phone take, a screen recording or an untextured 3D blockout into anime, claymation, a film grade or a photoreal render, while the blocking and the timing stay where they were.
One shot for a whole catalogue or cast: Keep the set, the lighting and the camera move, and swap who or what is in the frame: another character from a still, the product that actually shipped, a different car.
A motion guide for a new scene: Use a clip only for how it moves. The camera path and the choreography carry over, and reference stills decide what everything looks like.
A shot that ended too early: Continue a clip past its last frame when the edit needs one more beat and the original footage is long gone.
Which models do video to video here
Only the models that accept a clip appear in this table, and every value is read from the registry the composer enforces. The inputs column separates the two ways a clip can be used: as a reference the scene is re-rendered from, and, on Seedance 2.5 alone, as the source of a dedicated edit or extend job.
Model
Max resolution
Clip length
Audio
Inputs
On free credits
Starting cost
Seedance 2.5
1080p
4–30s
On/off toggle
Up to 10 reference clips · Edit and extend modes · Up to 30 reference images
Paid plan or pack
From 12 credits (4s · 480p)
Seedance 2.0
4K
4–15s
On/off toggle
Up to 3 reference clips · Up to 9 reference images
Paid plan or pack
From 9 credits (4s · 480p)
Seedance 2.0 Fast
1080p
4–15s
On/off toggle
Up to 3 reference clips · Up to 9 reference images
Paid plan or pack
From 7 credits (4s · 480p)
Seedance 2.0 Mini
720p
4–15s
On/off toggle
Up to 3 reference clips · Up to 9 reference images
Yes
From 5 credits (4s · 480p)
Kling O3
1080p
4–15s
On/off toggle
One reference clip
Paid plan or pack
From 10 credits (4s · 720p)
Kling 3.0 Motion Control
1080p
Follows the reference clip (3–30s)
No audio option
One reference clip
Paid plan or pack
From 9 credits (3s · 720p)
Kling 2.6 Motion Control
1080p
Follows the reference clip (3–30s)
No audio option
One reference clip
Paid plan or pack
From 6 credits (3s · 720p)
MiniMax H3
2K
5–15s
No audio option
Up to 3 reference clips · Up to 9 reference images
Paid plan or pack
From 23 credits (5s · 2K)
Wan 3.0
1080p
2–30s
On/off toggle
Up to 5 reference clips · Up to 10 reference images
Paid plan or pack
From 2 credits (2s · 480p)
Notes on this table
Of the entry models only Seedance 2.0 Mini accepts a clip, so a video-to-video test on free credits runs on Mini at up to 720p. Seedance 2.5, with the dedicated edit and extend modes, needs a paid plan or a credit pack, as do Kling O3, Kling 3.0 and 2.6 Motion Control, MiniMax H3 and Wan 3.0.
Output length follows the model, not the source. A 5-second source can become a 15-second Seedance 2.0 clip, and a 1080p source does not lift a 720p plan cap. The two Motion Control models are the exception: their output is as long as the reference clip.
Starting cost is the catalogue rate for the shortest, lowest-resolution prompt-only output the model offers, before plan discounts. A job with a clip attached is priced differently; the cost table below works four of them through in dollars.
Resolution cap by plan
The plan sets the ceiling and the model sets what is possible under it. A 4K-capable model still renders at 720p on free credits, and Seedance tops out at 1080p on every plan.
Plan
Seedance models
Other models
Free credits
720p
720p
Lite
720p
1080p
Pro
1080p
2K
Premium
4K
4K
Last updated
What your source clip has to be, model by model
These are the limits each model enforces on the clip you attach. They are worth a minute before you upload: 157 of the 379 video-to-video jobs that failed here between October 10 and November 9, 2026 broke one of them.
Model
Clip length
File size
Clips per job
On free credits
Seedance 2.5
2–30 s
Up to 200 MB
Up to 10
Paid plan or pack
Seedance 2.0
2–15 s
Up to 50 MB
Up to 3
Paid plan or pack
Seedance 2.0 Fast
2–15 s
Up to 50 MB
Up to 3
Paid plan or pack
Seedance 2.0 Mini
2–15 s
Up to 50 MB
Up to 3
Yes
Kling O3
3–10 s
Up to 200 MB
One
Paid plan or pack
Kling 3.0 Motion Control
3–30 s
Up to 100 MB
One + a still of the character
Paid plan or pack
Kling 2.6 Motion Control
3–30 s
Up to 100 MB
One + a still of the character
Paid plan or pack
MiniMax H3
2–15 s
Up to 50 MB
Up to 3
Paid plan or pack
Wan 3.0
1–15 s
Up to 100 MB
Up to 5
Paid plan or pack
Notes on this table
Format: MP4 or MOV on every model.
Length is the limit people hit most, in 95 of those 157 jobs. When several clips go into one job their combined length counts as well: 15 seconds in total on Wan 3.0 and 30 seconds on Seedance 2.5.
Picture size: the Seedance models reject a clip with fewer than 407,696 pixels per frame, so 854 × 480 passes and 640 × 480 does not, and they reject anything above 4K. Thirty-two jobs failed on the lower limit.
Frame rate: 60 frames per second is the ceiling on the Seedance models. Slow-motion phone footage at 120 or 240 fps has to be exported at 60 or lower first; 18 jobs failed on this.
Kling 3.0 and 2.6 Motion Control follow a reference clip of 3 to 30 seconds, or 3 to 10 when the character’s orientation is taken from the still rather than from the clip.
What a video-to-video job costs in dollars
On the Seedance models a job with a clip attached is billed on the seconds that go in plus the seconds that come out, at a per-second rate of its own. A 5-second clip turned into a 5-second result is therefore priced as 10 seconds. This table works four such jobs through at each plan’s price per credit, with the same calculation the generator bills with.
Plan and price
Seedance 2.0 Mini5 s clip → 5 s · 720p
Seedance 2.05 s clip → 5 s · 720p
Seedance 2.55 s clip → 5 s · 720p
Seedance 2.510 s clip → 10 s · 1080p
Lite, monthly$14.90/mo · 100 credits/mo
$2.0914 credits
$4.1728 credits
$6.1141 credits
Not on this plan
Lite, billed yearly$13/mo · 100 credits/mo
$1.8214 credits
$3.6428 credits
$5.3341 credits
Not on this plan
Pro, monthly$29/mo · 200 credits/mo
$1.6011 credits
$3.1922 credits
$4.6432 credits
$24.65170 credits
Pro, billed yearly$14.50/mo · 200 credits/mo
$0.8011 credits
$1.6022 credits
$2.3232 credits
$12.33170 credits
Premium, monthly$149/mo · 1,300 credits/mo
$1.1510 credits
$2.5222 credits
$3.6732 credits
$19.26168 credits
Premium, billed yearly$99/mo · 1,300 credits/mo
$0.7610 credits
$1.6822 credits
$2.4432 credits
$12.79168 credits
Starter credit pack$29.90 one-time · 200 credits
$2.2415 credits
$4.6331 credits
$6.8846 credits
$36.33243 credits
Notes on this table
Each dollar figure is the plan’s price per credit times the credits that job costs on that plan, for someone subscribing today. The credits a job costs differ by plan, which is why every row is computed on its own.
A shorter source is the cheapest saving available: every second trimmed from the clip comes off the bill, and a clip longer than the model accepts is the most common reason a job fails here.
Seedance 2.5’s edit mode is priced separately. The model chooses the output length itself, so the composer reserves credits for the longest case when you submit and returns the difference once the real length is known.
Sign-up credits are not in the table: they cover a first finished clip on Seedance 2.0 Mini at up to 720p, with no card on file. Lite caps Seedance at 720p, so the 1080p column is empty on Lite, and a failed or refused generation is not charged.
The composer shows the exact credit cost of the job you have set up before you press generate; if it ever disagrees with this table, the composer is right.
Timed on this site from the moment a job is submitted to the moment the finished clip is ready, across more than 700 video-to-video jobs that finished between October 10 and November 9, 2026. These are real waits, not progress-bar estimates, and every row has at least 30 finished jobs behind it.
Model
Setting picked most often
Median wait
3 in 4 done within
Seedance 2.0 Mini
4 s · 480p
2 min 53 s
3 min 19 s
Seedance 2.5
5 s · 480p
5 min 17 s
6 min 54 s
Seedance 2.0
10 s · 480p
5 min 27 s
6 min 44 s
Notes on this table
The wait covers everything between pressing generate and the clip appearing in your library: reading the source, the queue at the model provider, rendering and delivery. A job with a clip takes somewhat longer than the same model from a prompt alone, and longer outputs and 1080p stretch it further.
Jobs started in unlimited mode, which can wait in a queue by design, and draft previews are left out, so the figures describe an ordinary credit-billed run.
Free credits and Lite run one job at a time, Pro runs four in parallel and Premium eight, so on a paid plan several of these waits overlap.
MP4 or MOV, 2 to 15 seconds on Seedance 2.0 and up to 30 on Seedance 2.5. Stills added next join it as references, and on the Seedance models more clips can follow: up to three on Seedance 2.0 and ten on Seedance 2.5.
Step 2
Write the change, not the scene
Say what should be different and what must stay. “Replace the perfume in the gift box with the cream from the image” is the shape that works; a full scene description fights the source.
Step 3
Generate and compare
The result lands next to the original in your asset library with no watermark, so you can judge the change frame by frame and attach it back in for the next pass.
Four clips that started from another clip, each with the instruction that made it. Open one in the generator, attach your own footage, and the same model is already picked.
One-sentence product swap
Seedance 2.0 · 16:9 · 4s
Change the perfume in the video gift box to the face cream from the image.
Render the white model: upgrade the car body to a deep blue gradient metallic paint, with dark areas in navy blue and deep sapphire blue, naturally transitioning on lit surfaces to cobalt blue, bright blue, and electric blue, with fine metallic particles, clear-coat reflections, and flowing highlights. Upgrade the mechanical structure to cool gray metal, brushed metal, matte alloy, and sprayed steel materials. Upgrade the interior and seats to dark brown/walnut brown premium leather, with fine texture, slight wrinkles, and soft reflections. The glass and lamp covers should have a realistic transparent refractive texture …
Use @Video 1 as the white-model control input for the entire film, transforming the white-model previz into a refined, realistic chase scene with a strong commercial blockbuster feel, ultimately generating a 15-second highly realistic action film short. Use the reference images as subject appearance controls: @Image 1 controls the motorcycle’s model, color, structural proportions, and materials; @Image 2 controls the SUV’s model, color scheme, exterior silhouette, and details …
Use the white-model reference video as the sole guide for camera movement, pacing, framing, subject motion, and blocking. Combine it with stage keyframes to create a 30-second cinematic 3D animated short. Style: dreamy, warm, fairy-tale, childlike fantasy. Keep character designs consistent with the keyframes; do not alter appearances. Expressions should change with the scene …
The first example is the real pair that opens the examples at the top of this page, with its unedited instruction. The other two are prompt recipes for a restyle and an extension, written against Seedance 2.5’s modes so you can run them on a clip of your own.
01
Perfume out, face cream in
Before
A finished 5.5-second gift-box clip with a perfume bottle inside it, plus one still of the face cream that should be there instead. Two attachments, one sentence.
After
The same camera move, set and lighting with the cream in the box. Nothing else in the frame was mentioned, so nothing else moved. This is the first clip in the examples at the top of this page, and the shape of the instruction is the lesson: name the object to replace and the image it comes from.
Prompt
Change the perfume in the video gift box to the face cream from the image.
Settings
Model: Seedance 2.0
Inputs: one 5.5-second source clip plus one reference still
Entry: video to video, inferred from the attached clip
A handheld eight-second walkthrough of a furnished living room shot at noon: flat light, correct blocking, the kind of take you would keep if only the mood were different.
After
The same walk with the same timing, now at dusk: table lamps on, cool blue window light, longer shadows, with the furniture, layout and materials unchanged because the instruction told the model to leave them alone.
Prompt
Keep the camera path, framing and timing of the source clip exactly. Change the time of day to dusk: warm table lamps switched on, cool blue light through the windows, soft long shadows. Furniture, layout and materials stay the same.
Settings
Model: Seedance 2.5, paid plan or credit pack required
Inputs: one source clip, no stills
Duration: 8 seconds to match the source · Resolution: 720p on Lite, 1080p on Pro and Premium
Mode: edit, inferred from the instruction
Prompt recipe: no clip attached, paste it and run it.
03
Extend a shot that ended too early
Before
A six-second clip of a cyclist crossing a bridge at golden hour that cuts just as she reaches the far side, one beat before the edit needs it to.
After
The shot continues: she rides off the bridge into a narrow town street while the camera holds its pan at the same speed, and the light and grain match the original so the join does not read as a cut.
Prompt
Extend the video by 5 seconds. The cyclist rides off the far end of the bridge into a narrow town street, the camera keeps panning at the same speed, same golden-hour light and film grain, no cuts.
Settings
Model: Seedance 2.5, the only model with a dedicated extend mode
Inputs: the 6-second clip as the source
Duration: the seconds to add, 4 to 30; the result is the new segment only, to place after the original
Mode: extend, detected from the word “extend” in the instruction
Prompt recipe: no clip attached, paste it and run it.
Where video to video falls short, and why a job fails
Counted on the 1,109 video-to-video jobs run on this site between October 10 and November 9, 2026: 728 finished and 379 failed. Knowing which is which saves retries, and a failed or refused job is not charged.
Four in ten failures are the source file, not the idea
Of the 379 failed jobs, 157 never reached rendering because the clip itself was out of range: 95 were longer than the model accepts, 32 were below the minimum picture size, 18 ran above 60 frames per second, 7 were over the file-size cap and 5 had other dimension problems. Trim the clip, export it at 60 fps or lower and at 480p or above, and those failures do not happen. The table above has each model’s limits.
Copyright and trademark is the most common refusal
The content check stopped 176 jobs, and 72 of them named copyright or trademark: 42 in the picture the model was about to return, 18 in the soundtrack it generated and 12 in what was uploaded. Logos, brand names, franchise characters and recognisable songs are the triggers. Describe the look instead of naming it, keep logos out of the reference stills, and if the refusal names the soundtrack, switch audio off and run the job again.
Real people in the source clip
Twenty-one of the 1,109 jobs were refused because the uploaded clip or still showed a real person. That is under two in a hundred, but it is the refusal you cannot write your way around, since the trigger is the footage and not the prompt. Stylized, animated or clearly fictional subjects are the dependable choice.
A restyle redraws every frame
When the whole clip is re-rendered, nothing is copied pixel for pixel: small text, logos, fingers and background faces can come back slightly different even though the motion is right. When only one thing should change, use Seedance 2.5, whose edit mode changes the region you name and leaves the rest of the frame alone.
Long sources drift and cost more
The longer the clip, the more the model has to hold consistent, and on the Seedance models every second of source is billed. Keep the source to the seconds that matter, and cut a long sequence into shots that run one at a time with the same reference stills.
The result is new footage, not your file with a filter
Output length follows the model rather than the source, resolution follows your plan’s cap rather than the upload, and fast motion blur in the source is where artefacts appear first. A 4K source on free credits still returns 720p.
Video to video is generating a new clip from one you already have. The model reads the source for its motion, camera path, timing and layout, then follows your instruction for what should be different: a new style, a replaced object or person, a different setting, or more seconds after the last frame. It differs from text to video, where nothing exists yet, and from image to video, where a still becomes the first frame and the motion is invented. Here the entry is chosen for you: attach a clip and the composer switches to video to video.
What can I send in?
A source clip in MP4 or MOV, plus stills as references. Seedance 2.0, Fast and Mini take clips of 2 to 15 seconds up to 50 MB, and up to three per generation; Seedance 2.5 takes clips of 2 to 30 seconds up to 200 MB, up to ten alongside 30 reference images; Kling O3 takes one clip of 3 to 10 seconds up to 200 MB. The Seedance models also need at least 407,696 pixels per frame, so 854 × 480 or larger, and no more than 60 frames per second. Short sources read best, because the longer the clip runs the more the model has to hold consistent. Only upload footage you have the right to use.
Why did my video-to-video job fail?
Usually for one of two reasons, and neither is charged. Of 379 failed video-to-video jobs on this site between October 10 and November 9, 2026, 157 failed on the source file: too long for the model (95), too small a picture (32), above 60 frames per second (18) or over the size cap (7). Another 176 were stopped by the model’s content check, most often for copyright or trademark (72), followed by the generated picture (31), the uploaded clip (26) and a real person in the input (21). Trim and re-export the clip for the first kind; describe the look instead of naming a brand or character for the second.
Can I turn a video into anime, a cartoon or another style?
Yes. Attach the clip, say that the motion, framing and timing stay, and describe the target look with its materials, light and palette. The model redraws every frame in that style while the movement comes from your footage. Seedance 2.0 Mini does this on free credits at up to 720p, and Seedance 2.5 handles clips up to 30 seconds. Describe the style rather than naming a studio or a franchise, because copyright and trademark is the most common reason a job is refused here.
Can I replace a person in my video with another character?
Yes. Attach the clip and one still of the character who should appear instead, then write the swap as one sentence that names who goes out, who comes in from the image and that everything else stays. The movement, timing, camera and lighting are kept from the original. Free credits run this on Seedance 2.0 Mini at up to 720p; Seedance 2.5 does it as a localized edit on a paid plan or a credit pack. A source clip that shows real people can be refused by the content check, which happened to 21 of 1,109 jobs here, so a stylized or clearly fictional replacement is the dependable choice, and a refused generation is not charged. The task section above has the exact prompt.
Can I swap one object for another, such as a car or a product?
Yes, and it is the most reliable job on this page. One still of the replacement object, one sentence naming what to remove and which image it comes from, and nothing else in the sentence: the first clip in the examples at the top of the page replaced a perfume bottle with a face cream on Seedance 2.0 that way. The camera, set and lighting stay because you never asked them to move. Keep the source clip short and the reference still clean for the most stable result.
Can I change the background of a video without a green screen?
Yes. Say that the people, their movement and the camera path stay, then describe the new setting and its light. Seedance 2.5 treats this as a localized edit and keeps the subject stable while the environment is re-rendered; Seedance 2.0 and Mini re-render the whole scene from the clip, which suits a full restyle. A still of the new setting is optional and keeps a series of shots in the same place.
Can I add people or objects to footage I already have?
Yes. Name what appears, where it goes and what it does, add a reference still if it has to be a specific thing, and say the rest of the shot stays unchanged. One clear addition per run works far better than a crowd of small ones; stack further additions as separate runs on the result. Seedance 2.0 Mini handles this on free credits, and Seedance 2.5 gives a cleaner localized insert on a paid plan or a credit pack.
Is it free, is there a watermark, and do I need a card?
New accounts start with free credits and no card, sized for a first finished clip: two generation attempts and one completed video before a plan or a credit pack is needed. Nothing rendered for a signed-in account carries a watermark, on free credits or on a paid plan. One caveat specific to this entry: free credits cover the entry models, and of those only Seedance 2.0 Mini accepts a clip, so a free video-to-video test runs on Mini at up to 720p. Seedance 2.5, with its dedicated edit and extend modes, needs a paid plan or a credit pack.
Which model handles editing and extending, and do I pick a mode?
Seedance 2.5 is the one built for it: up to 30 seconds, up to 30 reference images and ten reference clips, and dedicated edit and extend modes that change one region or continue past the last frame instead of regenerating the whole shot. You never pick the mode; the instruction decides. Asking for a replacement runs an edit and asking for what happens next runs an extension. Seedance 2.0, Fast and Mini, Kling O3, MiniMax H3 and Wan 3.0 accept a clip and re-render the scene from it, which suits restyles and product swaps, and Kling 3.0 and 2.6 Motion Control are the specialists for transferring a performance from a clip onto a still.
How long and how sharp can the result be?
Output length follows the model, not the source. Seedance 2.5 renders 4 to 30 seconds, Seedance 2.0, Fast and Mini up to 15, Kling O3 up to 15, MiniMax H3 5 to 15, Wan 3.0 2 to 30, and Kling 3.0 and 2.6 Motion Control follow the reference clip, 3 to 30 seconds. Resolution is capped by plan first: 720p on free credits; then Seedance at 720p on Lite and 1080p on Pro and Premium, and other models at 1080p on Lite, 2K on Pro and 4K on Premium where the model allows it. A 1080p source does not raise a 720p cap.
What does a video-to-video generation cost?
The starting-cost column in the model table is the floor: the catalogue rate for the shortest prompt-only output, before plan discounts. On the Seedance models a job with a clip attached is billed on source seconds plus output seconds, at a per-second rate of its own, so a 5-second clip turned into a 5-second result is priced as 10 seconds. The cost table above converts four such jobs into dollars on every plan. Seedance 2.5’s edit mode reserves credits for the longest possible output and returns the difference when the real length is known, Kling 3.0 and 2.6 Motion Control are priced by the length of the reference clip, and the composer shows the exact figure before you press generate. Failed generations are not charged.
Is seeddance.io the official ByteDance site for Seedance?
No. SeedVideo AI at seeddance.io is an independent platform, not a ByteDance property. It runs the official Seedance model API alongside Kling, Veo, MiniMax and Wan, so the Seedance output you get here is produced by the model ByteDance publishes, but the workspace, the credits and the plans are ours. Model names belong to their respective owners, and nothing on this site is endorsed by them.
Can I use edited footage commercially?
Commercial use is included on Pro, Premium and the annual Lite plan, and not on free credits or the monthly Lite plan. The output is the same file at every tier; the licence is what changes. With this entry the input matters as much: the source clip and the reference stills must be yours to use, because a licence for the output does not create rights in footage you did not own.
How long does it take?
A few minutes. Measured on this site between October 10 and November 9, 2026, the median video-to-video job finished in 2 min 53 s on Seedance 2.0 Mini, 5 min 17 s on Seedance 2.5 and 5 min 27 s on Seedance 2.0, each at the setting people pick most often; the measured table above has the full figures. That is somewhat longer than a text prompt on the same model, because the source is read before rendering starts. The workspace shows an “about N min left” estimate while it works, and free credits and Lite run one job at a time where Pro runs four and Premium eight.
How do I keep the product or character consistent through the edit?
Give the model one clean reference of the thing that must stay fixed, name it in the instruction, and change nothing else in the sentence: “replace the perfume with the cream from the image” keeps the camera, set and lighting because you never asked for them to move. Keep the source short, avoid heavy motion blur, and for a run of variants reuse the same reference still on every generation. Seedance 2.5 takes up to 30 reference images, so a whole product line can travel with the clip.
Video to video here, or Runway Aleph or Kling, for a product swap?
All three can restyle a clip; the difference is how the change is scoped. Seedance 2.5’s localized edit changes the region you name and leaves the rest of the frame alone, and it does so with the same references you use everywhere else on this site, up to 30 images and ten clips. Kling O3 sits in the same picker, so you can run the identical source and instruction through both and compare frame by frame without leaving the workspace. Runway Aleph is not offered here, and we publish no benchmark against it; judge it on its own output.
Video, image or text: which entry fits
The three scene pages share one composer. What you attach decides the entry, and the model you pick decides the ceiling.
A clip is the right entry when the footage is almost right and only one thing in it has to change. When the shot does not exist at all, start from text to video and describe it. When you have a packshot, a poster or a portrait but no footage, start from image to video, where the still becomes the first frame and the prompt supplies the motion.
Within this entry the model matters more than anywhere else. Seedance 2.5 is the only model with dedicated edit and extend modes, and the only one that takes ten reference clips and 30 reference images, so localized fixes and extensions belong to it. Seedance 2.0 and its Fast and Mini tiers re-render a scene from a reference clip at a lower cost, which is enough for a restyle or a product swap. When a performance from one clip has to land on a still of someone else, use motion control: Kling 2.6 Motion Control does it at the lower credit rate, and Kling 3.0 Motion Control is the newer generation.
When the clip is right but ends too early, you do not need to change anything inside it: the AI video extender continues it from the last frame for 4 to 30 seconds on Seedance 2.5, with a worked example, the exact instruction and the cost per plan.
Where to go next
Take the clip further with another model, or check what your credits cover.