Up to 9 images
The default Seedance 2.0 workflow accepts multiple visual references for one generated video.
Preview
More AI Video Models
Compare compatible video models, then switch between reference-to-video, image-to-video, text-to-video, and video editing workflows.
FLUX 3 Video
Open FLUX 3 Video in AI Image Editor.
Gemini Omni
Open Gemini Omni in AI Image Editor.
HappyHorse
Open HappyHorse in AI Image Editor.
Seedance 2.5
Open Seedance 2.5 in AI Image Editor.
Seedance 2.0
Open Seedance 2.0 in AI Image Editor.
Wan 2.7
Open Wan 2.7 in AI Image Editor.
Veo 3.1
Open Veo 3.1 in AI Image Editor.
Reference to video AI
Turn a set of reference images into one controlled AI video
Upload the product, character, wardrobe, location, or visual style that the video should follow. Then describe the motion, camera, pacing, and details that must remain consistent across the generated clip.
Multiple reference imagesSubject and style continuityCamera and motion promptsSwitchable video models
Keep + animate
Separate the visual details that must stay recognizable from the action, camera, and atmosphere you want to add.
Short AI video
Choose supported resolution, ratio, and duration settings, then review the finished clip before publishing.
A reference-led workflow for controlled video generation
Reference-to-video starts with visual evidence instead of asking a prompt to define everything. Use the image set to anchor identity and art direction, while the prompt focuses on motion and camera behavior.
Guide one video with multiple images
Upload different angles, expressions, product views, wardrobe details, or location references so the model receives a clearer visual brief.
State what cannot change
Call out identity, silhouette, proportions, color, material, logo placement, clothing, furniture, or other details that must remain recognizable.
Direct motion and camera separately
Describe subject movement, environmental motion, shot size, camera path, lens character, pacing, and the final framing as distinct instructions.
Choose a model by reference needs
Start with Seedance 2.0, then compare compatible options such as Wan 2.7, Gemini Omni, HappyHorse, or Veo 3.1 where their current controls fit the job.
Prepare channel-ready ratios
Select landscape, portrait, or square output before generation so composition and camera movement fit the intended placement.
Review continuity before use
Inspect faces, hands, product geometry, logos, text, clothing, background objects, and motion continuity before downloading a final version.
Where multiple references improve an AI video brief
Use this workflow when one image does not explain enough: the subject has several important angles, the character needs a consistent wardrobe, or the campaign needs a repeatable visual language.
Preserve a product across a moving hero shot
Combine front, side, detail, packaging, and material references so the prompt can focus on the reveal, camera path, light, and final composition.
Use clean product angles
Choose references that clearly show shape, proportions, materials, controls, labels, and features that must not drift.
Describe a restrained camera move
Start with a simple orbit, push-in, pullback, or tabletop slide before asking for more complex choreography.
Inspect every product detail
Pause on key frames and check geometry, marks, reflections, packaging text, and contact shadows against the source set.
Keep a character recognizable through motion
Use portrait, profile, full-body, clothing, and palette references when identity and wardrobe continuity matter across a short scene.
Cover identity and wardrobe
Include clear references for face, hair, body proportions, clothing layers, accessories, and the intended visual treatment.
Limit the action first
Use one readable action and one camera move so the model has more capacity to preserve the character.
Check anatomy and continuity
Review face stability, hands, joints, fabric behavior, accessories, and background crossings throughout the clip.
Carry a campaign look into a short sequence
Reference an approved key visual, palette, surfaces, typography-safe area, lighting, and product treatment to make the generated clip feel related to the campaign.
Curate one visual language
Avoid contradictory references. Keep color, lighting, material, and composition signals deliberate and coherent.
Write the shot sequence
Describe the opening frame, main movement, transition, and final hero framing in a short ordered prompt.
Leave copy space intentionally
If the clip needs a headline or CTA later, reserve stable negative space instead of asking the video model to render final marketing text.
How to generate a video from reference images
01
Upload a coherent reference set
Choose images that explain the same subject or art direction, using clear angles and details without contradictory styles.
02
Write the keep and motion instructions
List the visual details to preserve, then describe subject action, environmental motion, camera path, pacing, and final frame.
03
Set output, generate, and inspect
Choose a compatible model, duration, ratio, and resolution. Generate the clip, then review continuity before downloading or refining the brief.
Reference to video AI FAQ
What is reference-to-video AI?
Reference-to-video AI generates a video from a prompt plus a set of uploaded visual references. The references guide subject identity, product appearance, wardrobe, palette, location, or style while the prompt directs motion and camera behavior.
How is reference-to-video different from image-to-video?
Image-to-video usually animates one source image. Reference-to-video can use several images to explain more angles, details, characters, products, or art-direction signals before generating the clip.
How many reference images can I upload?
The default Seedance 2.0 reference-to-video workflow accepts up to 9 images. Other compatible models can have different limits, so the current upload controls are the source of truth.
What should a reference-to-video prompt include?
Separate the prompt into details to preserve, subject action, environmental motion, camera movement, lighting, pacing, and final composition. Short, ordered instructions are easier to review and refine.
Can references guarantee perfect consistency?
No. References improve control but do not guarantee exact identity, product geometry, text, logos, anatomy, or continuity. Always inspect the full result before publishing or using it commercially.
Build a video from your reference set
Upload the visual brief, direct the motion, and generate a reviewable clip.
Multiple images InputKeep + animate Prompt