Transform one source video
Use the clip's motion, framing, timing, or performance as the foundation, then describe a new style, subject, environment, or treatment.
Transform existing footage or direct a new clip with references
This AI video to video generator brings compatible models into one editor. Start with a source video when you want to preserve broad movement, pacing, or composition while changing the style or scene. On supported models, add images, clips, or audio to guide identity, motion, camera direction, and sound.
Choose how your references guide the result
Create with multiple references
Use separate images, videos, and supported audio for identity, appearance, motion, camera direction, or sound when the selected model allows it.
One source video, two lighting directions
Use Video to Video when a clip already contains the movement you want to guide a new shot. These two edits use the forward-push clip below and change only the lighting sentence.
Seedance 2.0 Mini · one source video · requested 5 seconds, 480P, 16:9 · audio off · 61 credits per edit on October 4, 2026.
Warm afternoon light
Amber sunlight crosses the sofa and right wall, with stronger window-shaped shadows. The broad forward camera movement remains recognizable.
Prompt used
Edit @Video1. Keep the same room, furniture arrangement, camera path and five-second pacing. Change only the lighting to warm late-afternoon sunlight entering through the existing windows, with soft amber highlights and longer shadows. No new objects, no people, no cuts or text.
Cool overcast daylight
The windows and room take on cooler, more diffuse light. The candle stays warm and the main furniture arrangement remains recognizable.
Prompt used
Edit @Video1. Keep the same room, furniture arrangement, camera path and five-second pacing. Change only the lighting to cool overcast daylight through the existing windows, with soft diffuse blue-gray light and low-contrast shadows. No new objects, no people, no cuts or text.
Each downloaded result is 864 × 496 pixels and 5.04 seconds long.
Keep the same source clip when comparing light. The two edits retain the broad camera path while changing the atmosphere; small object shapes and reflections also change. For a closer revision, name the specific detail to retain and adjust one lighting instruction at a time.
Try a lighting editFour ways to guide a new video with source media
Start with source footage for a direct restyle, or give each optional reference a specific role and connect those roles in your prompt.
SOURCE VIDEO TRANSFORMATION
Restyle a street clip while reviewing its motion
Upload a short street clip and request a new night-time setting while retaining the forward camera move. Select a model that accepts source video and check its input limits and cost. Play the full result to compare movement, buildings, and transitions. The example illustrates a generated reinterpretation; it does not demonstrate a frame-perfect filter.
Example direction: Use the source for a slow forward camera move; change the street to a night scene with warm shop lighting.
CHARACTER CONSISTENCY
Place a referenced character in a new scene
Use a clear character image for appearance on a compatible reference model, then describe a new action and setting. Name the face, hair, and clothing to retain and review them across the clip. This example shows a character direction, not an identity lock. Use only references you have permission to upload.
Example direction: Use the image for face, hair, and clothing; show one gentle head turn in a softly lit room.
MOTION GUIDANCE
Use a performance clip to guide movement
Combine an appearance image with a motion clip when the selected model accepts both. State which input supplies appearance and which supplies the action or pacing. Review body proportions, gestures, and camera behavior throughout the output. Motion guidance can be interpreted or simplified; it does not guarantee identical choreography or timing.
Example direction: Use the video for movement and rhythm, the image for appearance, and keep one continuous shot.
MULTI-REFERENCE CREATION
Assign different roles to multiple references
On a compatible model, label each supplied image, video, or audio item with the tags shown in the editor. Give each one a role: character, setting, motion, or sound. Describe the final shot and inspect whether those roles carried through. Input types and counts vary, and audio references do not guarantee an exact voice match.
Example direction: Use the first image for the character, the second for the setting, and the clip for a slow camera move.What source media and prompts can guide
The exact controls depend on the selected model and its supported source media.
Source Video Restyling
Guide a new visual treatment with the source clip's broad framing and movement; compare the full result with the original.
Character Animation
Guide a character's appearance with an image on compatible models; check face, clothing, and proportions across the clip.
Motion Transfer
Name the action or rhythm to take from a motion reference; identical choreography and timing are not guaranteed.
Reviewable Output
Play the opening, middle, and closing moments to check identity, texture, transitions, and any generated sound.
Scene Reinterpretation
Describe a new environment around a referenced subject, then inspect whether its distinctive details remain recognizable.
Creative Control
Use the prompt and exposed model settings to guide action and camera behavior; exact paths and composition can vary.
Choose your next creative workflow
Start with an idea, a photo, or a clip. Pick what you want to create next.
Start with a prompt
Text to Image
Turn a written idea into a new image, from product concepts to illustrations.
Create an image
Start with a photo
Image to Image
Use your photo to guide a new style, setting, or visual edit.
Edit a photo
Start with a prompt
Text to Video
Describe a scene, action, and camera movement to create a short clip.
Create a video
Start with a still image
Image to Video
Animate a portrait, product photo, or artwork with a motion prompt.
Animate an image
Current workflow
Start with footage
Video to Video
Restyle a clip using its motion and composition as creative guidance.
Back to editorVideo-to-video requirements and supported controls
Use source footage when its broad motion, timing, composition, or performance should guide a new result. Available reference inputs, output controls, and credit estimates vary by selected model, so review the editor before each generation.
| Capability | What to expect | Basis |
|---|---|---|
| Primary input | Source footage plus a transformation prompt | Workflow definition |
| Transformation guidance | Broad motion, pacing, composition, or performance can guide the output; it is not a frame-perfect filter | Provider guidance |
| Optional references | Images, additional video, or audio only when the selected model exposes those inputs | Live editor controls |
| Limits and credit estimate | Use the upload validation, output choices, and estimate shown for the current model | Live editor controls |
Video-to-video inputs, prompts, and costs
Common questions about creating and transforming videos with reference media.
Can I generate an AI video for free without logging in?
Yes. Generate one video for free without logging in or signing up. Text to Video, Image to Video, and Video to Video share this trial. Guest videos use Seedance 2.0 Mini at 480P, up to 5 seconds. Duration and resolution controls require sign-in. Failed generations do not use the free try; further generations require credits.
What is an AI video to video generator?
AI Video to Video creates new footage guided by an uploaded clip and a transformation prompt. Use the source for motion, timing, composition, or performance, then describe the visual change. Select a model that accepts video, check its limits and credit estimate, and review the entire result. The output is a generative interpretation, not a frame-perfect filter.
Can I transform or restyle an existing video?
Yes. Choose a model that accepts source footage, upload a valid clip, and name the change and the details to retain. For example: Keep the forward camera move; change the street to a warmly lit night scene. Compare movement, subject shape, and transitions throughout the result. Generative restyling can also change details you asked to retain.
What should I use as a reference image?
Choose a sharp, unobstructed image that clearly shows the person, product, or style you want to guide the shot. Use one reference for each distinct role and name those roles with the tags displayed in the editor. Review faces, labels, and colors in the output. A reference guides appearance; it does not guarantee identical identity or products.
What does a reference video control?
A reference clip can guide broad movement, pacing, framing, or performance on models that accept it. State which action or camera behavior matters instead of asking the model to copy everything. Review the full generated clip against the source. Timing, choreography, and individual frames can change; the reference is guidance rather than a locked editing timeline.
Can I use reference images, videos and audio?
Images, video, and audio can be combined only when the selected model exposes those upload types. Choose the model first, check its upload limits, and assign each input a distinct role in the prompt. Inspect the resulting visuals and sound. Audio support does not guarantee voice identity, exact words, or synchronized performance.
How should I write the prompt?
Name each reference role first, then describe one main action, setting, camera direction, and the details to retain. For example: Use the image for appearance and the clip for movement; show the character walking through a garden in one slow tracking shot. Use the editor's reference tags and avoid conflicting instructions. Review before refining one direction.
How is Video to Video different from Image to Video?
Video to Video uses footage to guide motion or timing; Image to Video starts with a still image and generates new movement. Text to Video starts with a written brief. Choose the workflow by the source that must anchor the shot. All three generate new footage and can change subject details, so inspect the complete output.
Can the AI video editor use more than one reference?
Some models accept multiple references; the editor shows the available slots and permitted input types after selection. Give each image, clip, or audio source a different purpose and refer to its displayed tag. Avoid conflicting subjects and directions. More references do not guarantee better consistency, and unsupported files or combinations cannot be submitted.
How do input limits and generation credits work?
The selected model determines file types, sizes, duration limits, and reference counts; the uploader validates those requirements. Output duration, resolution, and billable reference media can affect the displayed credit estimate. Check the estimate after adding inputs or changing settings. One video trial is free without login, shared across all video workflows: Seedance 2.0 Mini, 480P, up to 5 seconds. Failed generations restore the trial. Further generations require sign-in and sufficient credits.
