Use reference images or a supported reference video to guide the identity, product details, objects, scene, or visual style of a new AI video. Reference-to-video is designed for guidance across a shot—not simply for choosing its first frame.
Input video
Sign-in is required to generate. Eligible new accounts currently receive signup credits, and the estimated credit cost is shown before submission.
References help communicate visual facts that are difficult to describe in text alone. Assign each image a clear purpose, keep references compatible, and use the prompt to define the new action and scene.
Guide facial features, hairstyle, clothing, silhouette, and other visible identity cues while placing a character into a new shot.
Try this direction
“Name the two or three identity details that matter most instead of asking for perfect consistency in every frame.”
Show the shape, material, color, and recognizable design elements of a product or object before directing a new campaign scene.
Try this direction
“Explicitly protect logos, dial layouts, packaging geometry, or other high-value product details.”
Guide color palette, lighting, texture, lens character, illustration language, and mood without forcing the output to start from the reference image.
Try this direction
“Describe which visual properties to borrow and request an original scene composition.”
Combine compatible angles or roles to communicate more information than a single image can provide.
Try this direction
“Keep all references consistent and identify each role: character, outfit, product, scene, or style.”
On supported inputs, use a short video to help communicate motion, camera rhythm, or scene behavior alongside image references.
Try this direction
“State whether the video is meant to guide movement, framing, or overall visual rhythm.”
Reuse a prepared reference set to explore new locations, actions, camera angles, and compositions for the same creative subject.
Try this direction
“Change one major variable at a time so you can evaluate reference fidelity more clearly.”
Use fewer, stronger references and tell the model why each one matters. Conflicting material can reduce consistency.
Upload clear, compatible images—or a supported reference video—and choose material that shows the identity, product, object, or style you want to guide.
State what each reference contributes, then describe the action, setting, camera, lighting, and the details that should remain recognizable.
Review the displayed model cost, generate online, compare the result with your highest-priority reference details, and download without a visible watermark.
Choose this workflow when visual identity, product design, or art direction should guide the whole shot instead of merely defining where it begins.
Explore new actions and environments while using the same prepared character appearance as visual guidance.
Stage recognizable products across studio, lifestyle, close-up, and social campaign shots.
Turn an art-direction board into moving scenes with a shared palette, texture, lighting, and mood.
Use a consistent visual reference set while testing a sequence of related scene ideas and camera setups.
The page opens directly in reference-to-video mode and keeps your input material, prompt, model choice, and displayed credit cost together.
Start in the correct mode instead of trying to force character or product guidance through a generic image-to-video workflow.
Choose available settings for duration, resolution, and aspect ratio while keeping the creative brief visible.
The current cost appears before submission, and available signup credits can be used toward eligible generations.
Download completed generations without a visible watermark for evaluation, editing, and presentation.
AI references guide generation rather than compositing the original asset frame by frame. Results can vary with reference quality, conflicts, motion, and shot complexity.
Default model
Wan 2.7 Reference
The page opens in the dedicated reference-to-video workflow.
Default setup
2s · 720p
Adjust available duration, resolution, and ratio controls in the composer.
Default cost
3 credits
Always confirm the live credit display before submitting because pricing can change.
Reference capacity
Up to 5 total assets
Use up to four compatible images plus one optional supported reference video.
Generation time
Varies
Processing depends on the model, selected settings, queue, and reference complexity.
Separate reference roles from the new scene instructions. Prioritize a few concrete identity or design cues instead of requesting abstract perfection.
Character reference
“Use the images as character references: preserve the round glasses, short black bob, yellow raincoat, and facial proportions. Create a new medium shot of the character waiting at a rainy bus stop, subtle expression, slow handheld camera.”
Product reference
“Use the references for the exact bottle silhouette, amber glass, cream label placement, and black cap. Create an original macro product shot on wet stone, slow orbit, crisp controlled highlights, no extra logo or text.”
Style reference
“Use the references only for their muted terracotta palette, paper texture, simplified geometric shadows, and editorial composition. Create an original animated street-market scene with a gentle lateral camera move.”
Multiple reference roles
“Reference 1 defines the character, reference 2 defines the blue jacket, and references 3–4 define the cabin environment and warm evening lighting. Create a new shot of the character opening the cabin door, preserve the named details.”
The correct mode depends on whether your source should guide visual identity, become the first frame, define both endpoints, or provide only a text brief.
Animate a single image when it should become the opening frame of the shot.
Define the exact opening and ending images for a directed transition.
Compare text, image, frame, reference, and audio workflows from one hub.
Explore a multimodal generation and editing workflow with image and video inputs.
Add compatible character, product, object, scene, or style references, describe their roles, and review the displayed credit cost before generation.
