Reference to video generator New
Generate the shot you want from reference images, videos and audio
Pure first-person flying camera perspective, no visible drone, no aircraft, no propellers. The camera races through a collapsing crystal canyon, skimming over glowing rivers, weaving between floating shards, then diving into a huge mirror portal. Final reveal: the canyon reflects into infinite copies of itself, exploding into a kaleidoscope universe of light. Seamless one-take, aggressive banking, realistic inertia, ultra-fast POV, high-end VFX, cinematic motion blur, volumetric glow, hyper-realistic.
Try this promptWhat is Reference to Video?
Show the model the character, the camera move and the sound you want. Reference mode takes reference images, reference videos and reference audio together: images set how people, products and places look, videos set how the camera and the subject move, and audio sets the music or the voice. One prompt ties them into a shot with sound: 4 to 15 seconds on the Seedance 2.0 family, up to 30 seconds on 2.5.
This page runs on ByteDance Seedance 2.5 / 2.0, locked to multi-reference mode. The Seedance 2.0 family takes up to 9 reference images, 3 reference videos and 3 audio clips, with each video or audio clip 2–15 s and 15 s in total per type. Seedance 2.5 raises that to 30 images, 10 videos and 10 audio clips, with each clip and each total up to 30 s. In the prompt, refer to them in upload order as @image1, @video1 and @audio1, for example: "the girl from @image1 walks down the street with the camera move from @video1, music from @audio1".
What you can make with Reference to Video
Images, video and audio together
Reference images set the look, reference videos set the motion and reference audio sets the sound, all combined in a single generation.
@ Image1 is the lead performer — preserve her exact ident
Reuse camera moves and motion
Upload a clip with a camera move or a dance, and the model follows its path and rhythm with your character or product in it.
Use the uploaded image as the exact visual reference
Consistent characters
Several reference images lock in faces, outfits and product details, which is what series and multi-shot stories need.
Create a 30-second fast-paced vertical video (9:16) set i
Refer to assets by number
Write @image1, @video1 and @audio1 in the prompt to tell the model what each asset is for.
A cinematic, high-contrast action video in a stark minima
Picture and sound are generated together; with reference audio the result follows its rhythm and voice as closely as it can.
You pay for the output length; with reference videos their length is also counted at a lower rate. The button shows the exact credits before you generate.
Made by the community
Seedance videos shared by creators. Hover to play, open one to see the full prompt and recreate it.
From idea to video in three steps
- 01Add your references
Drop in reference images (people, products, places), reference videos (camera moves, motion) and reference audio (music, voice). On the Seedance 2.0 family you need at least one image or one video, as audio can't be used on its own; Seedance 2.5 accepts audio alone. References can't contain real human faces.
- 02Write the prompt with numbers
Refer to assets in upload order as @image1, @video1 and @audio1, and say who does what, how the camera moves and how the sound is used. Then pick the version, duration, ratio and resolution.
- 03Generate and refine
Hit Generate and follow progress on the right. If it isn't right, swap a reference video or change how one asset is used in the prompt, and run again.
- Give every asset a clear job in the prompt, such as "camera move from @video1" or "wearing the outfit from @image2". Assets you don't mention may be ignored.
- Short, clean reference videos work best: one clear camera move or action is easier to reproduce than a fast-cut montage.
- Use reference images with a clear subject and a simple background. Don't use photos of real people, as real faces are blocked; create the character with Seedream first, or use a video you generated with Seedance here as the reference.
- Matching the output ratio to the reference video gives the best results; if you change it, describe the framing in the prompt.
Try these prompts
Click one to drop it into the generator above, then tweak a few words.
The product from @image1 sits in the centre of the table, the camera pushes in and orbits exactly like @video1, soft studio light, premium commercial look
Try itSame dance, your characterThe character from @image1 on a city rooftop performs the full dance from @video1 with the same rhythm, locked medium shot, warm evening light
Try itTalking-head voice-overThe woman from @image1 talks to camera, lip movement and tone following @audio1, bright cafe background, shallow depth of field, natural handheld feel
Try itTwo characters, one shotThe man from @image1 and the woman from @image2 meet on a neon street on a rainy night, the camera slowly pushes from a wide shot to a two-shot, cinematic grade
Try itBuilt for these creators
Put the camera moves, transitions and dances of trending clips on your own characters, fast.
Pair one product photo with a reference ad and get a product video with the same pacing and camera language.
Character sheets keep the look locked while reference videos set the blocking, for consistent multi-shot sequences.
Reference audio sets the beat and the voice and the picture follows, ideal for talking heads, MVs and beat-synced edits.
Reference to Video FAQ

Try Reference to Video now
No prompt skills needed. Write a sentence, pick a model and generate; the first images are on us.

