What is reference to video?

An AI video workflow that conditions generation on images, video, and/or audio you provide, instead of text alone.

Reference to video vs motion control?

Motion control copies a performance onto a character image. Reference to video uses media as style, identity, and scene guides for a new shot.

Can I include dialogue in the generate?

Some models emit audio. For accurate mouths, generate picture, then run lip sync with a script or clone.

Start with one identity still plus a prompt. Add more only when the first take missed a constraint.

How many credits does a reference to video clip cost?

The first 30 seconds of any generation, including reference to video, cost 150 credits ($1.50). Every extra second costs 5 credits ($0.05).

AI Video · reference to video

Reference to Video

Reference to video is the multimodal slot. When a prompt is not precise enough, you @-mention stills, a motion clip, or a sound bed so the model keeps the parts you already like. It is how cinematic generators on this site take four inputs at once. It is not motion control (which copies a performance 1:1 onto a character still) and not ad clone (which copies an ad framework onto a SKU).

Make a clip

Generate from references

150 credits · $1.50

Video model

Shape

Quality

480p · 720p · 1080p · 4K

Length

4–15 seconds

Native audio

Reference still

Character library

Pick a saved look or a public character, then generate. You can still upload your own still.

Open Character Library

—

No avatar selected yet

Public library

52 looks

Example · Seedance 2.0

Prompt

Each model has its own length and resolution caps. Seconds and quality chips update when you switch Seedance 2.0, Seedance 2.5, Veo 3.1, or Wan 2.5. Paste a Wavespeed key on Admin → Models.

See an example

Seedance 2.0 image-to-video example from Wavespeed. Sample generated by this tool’s model. Your clip uses your photo, script, or prompt.

Film lighting used as an AI video prompt reference

Camera setup inspiring an image-to-video shot

Why people use reference to video

Multiple references

Face from still A, room from still B, camera from a clip — say what each reference is for.

Audio as a reference

Some models can follow a bed or a line. Lip sync still wins when words must land.

Character consistency

Reuse the same avatar still as a reference instead of hoping the prompt remembers them.

Model switch in one studio

Try a draft model, then a heavier one, on the same reference set.

Three simple steps

Step 1
Add the references
Image, video, and optional audio. Label them in the prompt.
Step 2
Write what should change
Keep identity, change location — be explicit.
Step 3
Generate, then mouth-pass if it speaks
Video lip sync or talking avatar on the take.

What this clip costs in credits

Every reference to video job uses the same meter: 150 credits ($1.50) for the first 30 seconds, then 5 credits ($0.05) per extra second. Subscriptions price credits at $0.01; one-time packs are $0.08 each.

Duration	credits	Generation price
15s	150	$1.50
30s	150	$1.50
45s	225	$2.25
60s	300	$3.00
90s	450	$4.50
120s	600	$6.00

First 30 seconds = 150 credits ($1.50). Plan credits roll over while you stay subscribed.

Reference to Video

See an example