Reference to Video AI

Turn reference images into a short video that keeps the same characters, products, and places. Add video or audio on routes that accept them, assign each file, and download a private video.

Owned reference packs
Character · Product · Scene
16:9 · 9:16 · 1:1
Private videos

Reference to Video examples

Reference packs and the scenes they point toward. Each demo shows the images used and the resulting video.

Reference to Video example: three friends taking a daylight selfie in front of a stone cathedral, then looking at one another

Cathedral day selfie

Prompt

Three friends from @Image1 take a daylight selfie in front of a stone cathedral, phone held out and smiles toward the lens, then lower the phone and look at one another. Slight handheld sway and street ambience.

Reference to Video use cases

Keep a person in a new scene, keep a product in a new setting, put two people in one place, or combine separate images into one shot.

Character

Keep the person, change the scene

Hold the person from one image, place them in the setting from the other images, and give them a new action.

Reference to Video example: a red-haired knight in silver armor

Reference to Video example: a winged white lion with a saddle

Reference to Video example: a field of mixed wildflowers under a blue sky

Product

Keep the product, change the setting

Start from the product image, then put that same item with the person in the room from the other files.

Reference to Video example: a man in a worn shirt and suspenders

Reference to Video example: a tall bacon cheeseburger

Reference to Video example: a wooden cabin table by a window

Social

Put two people in one shot

Use one image per person and keep the bar from a third. The second person walks in from the doorway, and the room stays recognizable.

Reference to Video example: a man in a dark hat and leather coat

Reference to Video example: a woman in a silver dress

Reference to Video example: a dim bar with stools, bottles, and a pool table

Remix

Combine separate images into one shot

Point the prompt at a character, a vehicle, a place, and an object so the shot borrows from each file.

Reference to Video example: a desert highway with a Route 66 sign and a gas station

Reference to Video example: an orange cat in a cowboy hat and denim vest

Reference to Video example: a dry tumbleweed

Reference to Video example: a red convertible with a black soft top

What Reference to Video can do

Keep the images you upload recognizable, guide motion from a video when a route accepts it, and guide sound from an audio file when a route accepts it.

Keep the references recognizable

The images you upload are what the new video should hold onto. Ask for a new action or camera, and those references should stay recognizable. Details you leave out can still change.

Guide motion from a video

On routes that accept video, add a short clip so the camera or the action has a reference. Routes that only take images stay limited to images.

Guide sound from an audio file

On routes that accept audio, add a sound file so voice, music, or ambience has a reference. Some routes still need an image or video alongside that audio.

40 welcome credits
Exact cost before you run
Failed runs cost no credits

Reference to Video vs Image to Video

Use a reference pack when more than one image has to survive into the video. Use Image to Video when one frame, or a start and end frame, is enough.

Reference to Video vs Image to Video
Compare	Reference to Video	Image to Video
What you upload	Several images, plus video or audio on routes that accept them	One start image, and an optional end image on supported routes
What the video does	A new scene that tries to keep those references together	Camera and motion built from the start frame
How you direct it	A scene prompt plus @Image, @Video, and @Audio tags	A motion prompt for the start frame

Reference to Video models

These routes accept a reference pack. Limits differ by model. The generator shows the controls for the route you pick.

Image, video, and audio

Seedance 2.0

Image, video, and audio

Up to 9 images, 3 videos, and 3 audio files, 12 files total. Output is 4–15 seconds. Audio needs at least one image or video.

Larger packs

Seedance 2.5

Larger packs

Up to 30 images, 10 videos, and 10 audio files, 50 files total. Output is 4–30 seconds at 480p–1080p. Audio-only packs are allowed.

Mixed media packs

Wan 3.0

Mixed media packs

Up to 10 images, 5 videos, and 5 audio files. Output is 2–30 seconds.

Character packs

MiniMax H3

Character packs

Up to 9 images, 3 videos, and 3 audio files, 12 files total.

Higher-end packs

MiniMax H3 Max

Higher-end packs

The same 9-image, 3-video, 3-audio pack, 12 files total, for a higher-end take.

Image packs only

Grok Imagine Video 1.5

Image packs only

1–7 images at 480p or 720p. No video or audio references. Runs are at least 6 seconds.

How to generate AI video from reference images

Three steps in the generator on this page. Nothing is submitted until you confirm the quote.

1
1. Add the references you own
Upload images you are allowed to use. On routes that accept them, add a short video or audio file as well.
Upload images you are allowed to use. On routes that accept them, add a short video or audio file as well.
2
2. Describe the scene
Write the action and camera, then assign each file with @Image1, @Video1, or @Audio1. Pick format and model. The quote updates with that choice.
Write the action and camera, then assign each file with @Image1, @Video1, or @Audio1. Pick format and model. The quote updates with that choice.
3
3. Generate and download
Confirm the quote, generate a private video, inspect it, then download or open it in Create.
Confirm the quote, generate a private video, inspect it, then download or open it in Create.

Try Reference to Video AI ↑

Reference to Video Pricing

Seedance 2.0 starts at 33 credits for a 5-second 480p video. Other reference routes quote separately. Confirm the reservation before a private video.

Explore more AI tools

Animate one image, start from a written scene, or copy a performance onto an image.

Generate

Image to Video

Animate one owned image when one start image is enough.

Generate

Text to Video

Write the scene from scratch when you do not have references yet.

Edit

Motion Control

Copy a performance video onto a character image.

Reference to Video AI FAQs

What is Reference to Video?

Reference to Video generates a short video from reference files you own. You can start with several images and, on routes that accept them, add video or audio. It is not Image to Video from one start image, not Motion Control copying a performance, and not Video to Video restyling a source video.

How do I generate AI video from reference images?

Three steps in the generator on this page. 1) Add the references you own: upload images, and video or audio when the selected route accepts them. 2) Describe the scene: write the action and camera, assign each file with @Image1, @Video1, or @Audio1, then pick format and model. 3) Generate and download: confirm the quote, generate a private video, inspect it, then download or open it in Create.

Is MixVio Reference to Video free?

An account is required. New accounts receive 40 welcome credits valid for 7 days. Reference to Video is available on the Free plan when your balance covers the quote. Failed runs cost no credits.

How do credits work?

Credits are quoted from the selected model, duration, and resolution before you generate. The default 5-second 480p Seedance 2.0 video costs 33 credits. Other reference routes quote separately. Failed or cancelled tasks release the reservation.

What kinds of prompts work best?

Name the action and camera, then assign each uploaded file with @Image1, @Video1, or @Audio1. Use the number for that file’s slot so the person, product, place, motion, or sound stays assigned. The empty composer asks for the scene and motion; the tags appear once files are attached.

Will every face, product, and place stay exact?

No. Identity, logos, packaging, and fine place details can drift. Treat the demos as creative direction, review every video before you publish, and replace a reference or rewrite the prompt when something important slips.

What if I am not satisfied with the result?

Adjust the prompt, swap a reference, or switch model, then generate again. Each new run shows its quote before submit. Failed runs cost no credits.

Which models accept reference packs?

Seedance 2.0, Seedance 2.5, Wan 3.0, MiniMax H3, MiniMax H3 Max, and Grok Imagine Video 1.5 accept reference packs, with different file limits. Kling, Veo, FLUX, and PixVerse routes on MixVio stay on Image to Video: one start frame, and an end frame only where that route offers it. Compare the pack routes in the Models section.

What input limits apply?

Seedance 2.0 allows 9 images, 3 videos, and 3 audio files, 12 files total, for a 4–15 second video. Audio on that route needs at least one image or video. Seedance 2.5 allows 30 images, 10 videos, and 10 audio files, 50 files total, for 4–30 seconds at 480p–1080p, including audio-only packs. Wan 3.0 allows 10 images, 5 videos, and 5 audio files, for 2–30 seconds. MiniMax H3 and MiniMax H3 Max allow 9 images, 3 videos, and 3 audio files, 12 files total. Grok Imagine Video 1.5 allows 1–7 images at 480p or 720p, with no video or audio references, and runs of at least 6 seconds. The generator enforces the limit for the model you pick.

Can I use the videos for commercial projects?

Paid-plan outputs may be used commercially, subject to the Terms of Service and each model provider’s usage rules. You must own or be authorized to use every uploaded reference, and have rights to every face, brand, and prompt you submit. Free-plan outputs are for evaluation and are not for commercial publishing.

Are uploads and results private?

Yes. Reference uploads, prompts, and finished videos stay as private workspace assets. Signed URLs, asset IDs, and private prompts are not added to public page URLs or analytics.

How is this different from Image to Video?

Image to Video animates one start image, with an optional end image on routes that support a last frame. Reference to Video keeps a pack of images together, and can add video or audio on routes that accept them. One or two images without a pack are treated as a start frame or a start-and-end pair.

Open Image to Video →

How is this different from Motion Control and Video to Video?

Motion Control copies the motion of a performance video onto one character image. Video to Video restyles or recasts a source video you already have. Reference to Video starts a new scene from the reference pack instead of copying one performance or restyling one take.

Open Motion Control →

Add the references.
Generate the scene.

Upload references you own, assign each file, confirm the quote, and download a private video.

Start Reference to Video