Vidu Q4

Guide · Reference to video

Vidu Q4 reference to video, step by step

Reference to video is the Vidu Q4 route for recurring characters, branded products and fixed locations. You upload up to 12 images and up to 3 short voice clips, tag them in the prompt, and Vidu Q4 builds a new scene that keeps them consistent.

Start a reference video

In short

  • Up to 12 PNG, JPG or WebP images and 3 MP3 voice clips of 3 to 12 seconds.
  • At least one image is required; there is no text-only mode.
  • Five aspect ratios: 16:9, 9:16, 1:1, 4:3 and 3:4.
  • Sound is optional, and always on when a voice clip is attached.

What to upload

Use one clean image per subject: a person on a plain background, the product on white, an empty location. Mixed collages make it harder for the model to tell subjects apart. Only upload images and voices you have the right to use.

  • Characters: a full-body or half-body photo with the outfit you want kept.
  • Products: a sharp product shot showing the side that should appear.
  • Locations: the place without people, in the light you want.
  • Voices: an MP3 of the speaker, 3 to 12 seconds, with little background noise.

How to run it

  1. Add two or more images, or a voice clip. The studio switches from image to video to reference to video automatically. Add them in a deliberate order.
  2. Mind the order. The first image is @Image1 (Vidu’s [@reference_image_1]), the second @Image2, and so on. Voice clips are @Audio1 to @Audio3.
  3. Write the scene with tags. Put each tag where that subject acts, then add camera and sound.
  4. Pick ratio, length and resolution. Any length from 3 to 16 seconds; 2 credits per second at 540p up to 10 at 4K.
  5. Review and iterate. If a subject drifts, make its reference cleaner or name it more explicitly in the action.

Tested example: a three-reference sneaker ad

We uploaded a studio photo of a woman, a product shot of a sneaker and an empty rooftop court, then tagged them in a 5-second 9:16 prompt at 720p. The result is the vertical ad on our homepage: the same face and jacket, the same sneaker, on the same court, across several cuts.

Prompt
[@reference_image_1] sits on a bench on [@reference_image_3] and laces up [@reference_image_2]. She stands, dribbles a basketball and makes a jump shot at golden hour. She turns to the camera and says: "New season, new kicks." Handheld camera, quick cuts. Sound: ball bouncing, sneakers squeaking, city ambience.

Reference to video or image to video?

Use image to video when you already have the exact first frame and want it animated. Use reference to video when you need a new composition that still contains your subjects, or several subjects that never appeared together in one photo.

Frequently asked questions

How many references does Vidu Q4 support?

ShengShu announces up to 15 image references and 3 audio references. The fal route used here accepts 12 images and 3 MP3 voice clips.

Do I have to tag references?

A single untagged reference was still used in our test, but with more than one reference tags such as @Image1 make each subject’s role explicit and are strongly recommended.

Can I use only a voice clip?

The model accepts a voice clip alone, but this studio requires at least one image so that every voice belongs to a visible subject.

What if a render fails?

Credits are reserved when the job starts and returned automatically when a failure is confirmed. A job that is still being checked stays pending until it is reconciled.

Sources

Updated 2026-10-08. ViduQ4.org is an independent site, not affiliated with or endorsed by ShengShu Technology or Vidu. Vidu is a trademark of its owner. Videos are generated with the public Vidu Q4 image-to-video and reference-to-video routes on fal, as shown in the studio. Every example on this page was made with those routes on this site.