Reference to Video AI
Reference to Video AI for a more specific scene
Build a new scene with Reference to Video AI using images, video and audio supported by your chosen model. Give each upload a role: a character to recognize, a place to recreate, a movement to follow or a sound to guide the atmosphere.
Create with references
Reference to Video AI: give every asset a role
Multimodal video generation is useful when words alone leave too much open. Show the appearance, motion or sound you mean, then explain how the references belong together.
01 / IMAGE
A subject or a setting
Use an image to show a character, product, environment or visual treatment. Identify which details should influence the new scene.
02 / VIDEO
Movement and pacing
A video reference can communicate action or camera direction. Specify whether the model should follow the movement or borrow the overall feel.
03 / AUDIO
Sound direction
Supported audio references can guide the sound of a clip. State the intended role and check that the model accepts your combination of inputs.
Direct with @ references
Make your Reference to Video AI prompt unambiguous
“Use these references” leaves the model to guess what matters. Separate the jobs: this image supplies the person, that image supplies the landscape, and the video supplies the action.
If two references disagree, say which takes priority. A daylight location photo and a night scene can work together when you explicitly ask to retain the place but change the lighting. Fewer relevant inputs often make the brief easier to understand.
How to create a clip with Reference to Video AI
Choose compatible media
Select a model, then check its supported image, video and audio inputs. Add only references that contribute to the scene you want.
Assign each reference
Insert media with @ and explain the role of every asset. Add the action, camera direction and atmosphere for the new clip.
Generate and compare
Review settings and credits, generate, then inspect Your videos. Compare the output with each reference’s intended role before downloading.
Choose Reference to Video AI when the scene needs several kinds of direction
A reference-led workflow creates a new clip. Choose a more specific tool when your task is to revise footage or transfer a single performance.
Character and location concepts
Combine a character image with a place reference and describe their action. Review appearance across the full clip; references guide identity but do not guarantee exact preservation.
An existing video that needs a change
Use AI Video Editing when the source footage itself is the subject of the edit, rather than creative direction for a new shot.
A character following a performance
Use AI Motion Control for a dedicated pairing of one character image and one driving video.
Reference to Video AI: frequently asked questions
It is video generation guided by supplied media and a prompt. This page can accept image, video and audio references when the chosen model supports them. The output is a newly generated clip, not a stitched collection of uploads.
Image to Video focuses on still-image inputs. Reference to Video also exposes supported video and audio inputs, making it useful when motion or sound is part of your direction.
Only some models support audio-only reference input. Check the selected model and any upload validation shown in the generator. If the combination is unsupported, add a suitable visual reference or choose another available model.
Use AI Video Editing when the source footage is the main thing you want to change. Use Motion Control when your specific goal is to transfer a recorded performance to a character image.
Make your next idea visible