Reference to Video: Combine Images, Clips and Audio into One AI Video

Most image-to-video tools take one picture and animate it. That works for a single shot, but as soon as you want this character in that place, moving like this clip, with this music, one picture is not enough.

Reference to Video lets you add several references at once — still images, short video clips and audio — and say in the prompt which one is which. The model uses them together: a face and outfit from one image, a location from another, the camera move from a clip, the rhythm from a track. It runs on Seedance 2.5 and makes videos up to 30 seconds long, up to 1080p.

• Images: JPG, PNG or WebP, up to 30.
• Video clips: MP4 or MOV, 2–30 seconds each, up to 10.
• Audio: MP3 or WAV, 2–30 seconds each, up to 10.
• Up to 50 references in total, and the reference videos together can be up to 30 seconds.

You need at least one reference. Images alone are fine; clips and audio are optional.

Each reference gets a tag in the order you add it: @Image1, @Image2, @Video1, @Audio1 and so on. Press "Insert" next to a reference to drop its tag into the prompt, then write what should happen.

• @Image1 walks down a rainy street at night. The camera follows from behind. Music: @Audio1
• The woman in @Image1, in her 30s, sits in the café from @Image2 and turns to the window. Camera move like @Video1.
• The product in @Image1 rotates slowly on the table from @Image2, soft morning light, 8 seconds.

Say what each reference is for: "the face from @Image1", "the place from @Image2", "the camera move from @Video1". A reference that the prompt never mentions is used much more loosely.

• Make a new video from references — the default. Everything you add is material for a new shot.
• Remake a reference video (@Video1) — keep the timing and movement of your clip but change what is in it: the outfit, the place, the time of day, the style.
• Continue a reference video (@Video1) — make what happens next, picking up from your clip.

Remaking and continuing need one reference video.

Choose 4 to 30 seconds and 480p, 720p or 1080p. Credits depend on the length and resolution, and reference clips count too, because the model has to process them:

• 5 seconds from images only: 300 credits at 480p, 650 at 720p, 1,625 at 1080p.
• 8 seconds at 720p from images only: 1,040 credits.
• 10 seconds at 720p with an 8-second reference clip: 1,405 credits.
• Remaking a 12-second clip at 720p: 1,875 credits (a remake comes out as long as the clip).

The exact amount is shown on the button before you create anything. If a video fails, the credits come back automatically. Reference to Video is available on the Pro plan and above.

• Start with images only. It is the cheapest way to check that the model understands your references. Add clips and audio once the look is right.
• Use short, clean clips. A 3–5 second clip with one clear camera move works better than a long clip where many things happen.
• Keep the cast small. One or two people per shot keeps faces consistent.
• Only use material you have the right to use. When you add video or audio, you confirm that you have the rights and the consent of anyone who appears or speaks in it. Don't use other people's footage, music or voices without permission.

Ready to try? Open Reference to Video. For a single picture, image-to-video is simpler and the first one is free.