Reference to Video AI: Keep Characters Consistent Using Reference Images
Upload photos of a character, product or place, add a motion clip or sound if you like, and get a brand-new shot that keeps them looking the same. Switch between Seedance, Kling V4 and Grok on one credit balance, and get your credits back if a video fails.
- Character and product references, up to 30 images
- Reference clips for motion, audio for sound
- 3 to 30 seconds, up to 1080p, with sound
- Free credits to start
What Is Reference to Video AI?
Reference to video AI uses your reference images, video clips or audio as a guide. The AI analyzes who or what is in each file and creates a brand-new shot, so your characters, products and style stay consistent across scenes instead of drifting from clip to clip.
Character and Product Lock
Upload a reference image of a person, pet, product or place. The AI keeps the face, outfit, logo and material details, even from different angles, in new lighting or in a whole new setting.
Motion and Camera Control
Upload a reference clip, such as someone dancing or a slow camera push in. The AI copies the movement, rhythm and camera movement onto your own character.
Sound as a Reference
Add a voice line, song or sound effect as an audio reference. The video follows its rhythm and mood, so motion and sound line up in one shot.
Mix References in One Prompt
Combine character images, motion clips and audio references in a single video. Each upload gets a tag like @image1 or @video1, so your prompt can say exactly which file plays which role.
Reference to Video vs Image to Video
Both start from your images, but they use them in different ways. Image to video animates one picture from its first frame. Reference to video reads identity, style and material from your files and builds a new shot around them.
| Feature | Image to Video | Reference to Video |
|---|---|---|
| What your image becomes | The exact first frame of the video | A visual reference for who or what appears |
| What you can upload | One start image, plus an optional end frame | Up to 30 images, plus motion clips and audio |
| Composition | Keeps the framing of your photo | New angles, lighting and settings from your prompt |
| Several subjects | Only what is already in the photo | People, products and places from different images in one shot |
| Best for | Bringing one photo to life with camera movement | Recurring characters, product ads and multi-shot stories |
When is reference to video the wrong tool? If you need the exact composition of one photo, or fixed start and end frames, use image to video. If you have no images at all, start with text to video.
What Can You Create with Reference to Video AI?
Five ways people use reference images to keep their videos consistent.
Recurring Characters for Series and Stories
Keep the same face and outfit in every episode of an AI short drama, comic adaptation or YouTube series. Reuse one character turnaround sheet, and every new clip looks like the same person.
Upload one turnaround sheet per character: @image1 is Mira and @image2 is her fox. She runs through a rainy neon alley as the fox jumps onto her shoulder.
Product Ads for Paid Social
Put your product image into any scene without a photo shoot. The logo, label and shape stay consistent, so the same product can star in a kitchen, a gym bag or a beach ad.
Upload a product image on a plain background and a mood reference: the bottle from @image1 rises out of black water as flowers like @image2 bloom around it, label sharp at the end.
Virtual Try-On for Fashion and Ecommerce
Show how clothes, shoes or accessories look on a model without a photo shoot. The AI dresses the model in your items and keeps the fabric, color and logo details consistent as they walk and turn.
Upload a model and one image per item: @image1 wears the jacket from @image2, the sneakers from @image3 and the sunglasses from @image4, walking across a rainy Tokyo crosswalk at night.
Social Media Content for TikTok and Reels
Join a dance or meme trend with your own character. Upload the trend video as a reference clip, and the AI copies the moves, timing and camera while your character keeps their look.
Upload a character image, a short dance clip and a beat: the statue from @image1 performs the dance from @video1 to the beat of @audio1, then freezes back into place.
Brand Mascots and Logos
Animate a mascot or brand character and keep it consistent across every campaign, from launch teasers to holiday posts.
Upload a turnaround sheet of your mascot plus your logo: the mascot from @image1 pops out of the logo in @image2 and throws flour like snow.
How to Use Reference to Video AI in 3 Steps
From upload to preview in a few minutes. No video editing needed.
Upload Your Reference Images
Upload 1–3 reference images to start, one for each character, product or place. Add a reference clip for motion or an audio file for sound if you need them.
Write a Prompt That Names Each Reference
Type @ to tag a file, then describe the action and the camera movement: @image1 hands @image2 to a customer in @image3, slow push in.
Generate, Preview and Refine
Pick a model, aspect ratio and length, then click generate. Preview the result, refine the prompt if something drifts, and download your MP4.
How to Write a Prompt with Reference Images
Give every reference a job. Name it with its @ tag, say what it does, then describe the motion and camera. The tags work the same way on every model in the generator.
| What you want | What to upload | Example prompt |
|---|---|---|
| Your character in a new scene | A character turnaround sheet | @image1 walks through a snowy market at night and looks up at the lanterns, camera follows from behind |
| A product in someone's hands | A product image and a person | @image2 lifts @image1 from the table and smiles at the camera, soft daylight, close up |
| Copy a dance or movement | A character image and a motion clip | @image1 performs the dance from @video1 on a rooftop at sunset, same timing and camera |
| Move in time with music | A character image and an audio file | @image1 dances to the beat of @audio1 in a neon-lit studio |
| Several subjects together | One image per subject, plus a place | @image1 and @image2 sit on the sofa in @image3 and laugh at a movie |
Leave looks out of the prompt. The references already show them, so spend your words on the action, the mood and the camera.
What Makes a Good Reference Image or Clip?
The AI can only keep what it can see clearly. Better references matter more than a longer prompt.
| Reference | Do this | Why it helps |
|---|---|---|
| Character images | One turnaround sheet per character, with front, side and back views in one image | The AI sees the face and outfit from every side, so they stay stable when the character turns |
| Sheet layout | Views at the same size on a plain background, with no labels or text | Labels and clutter can leak into the video |
| Product image | A sharp, straight-on shot on a plain background with readable logos | Text and logos are the first details to warp |
| Reference clip | A short clip with one moving subject and steady framing | Clear motion is easier to copy onto your character |
| Audio reference | A clean voice line or song without background noise | The video gets a clear rhythm and mood to follow |
| How many | Only the references the shot needs, usually 1–3 | Extra images raise the chance of mixed up subjects |
Images: JPG, PNG or WEBP up to 10 MB. Reference clips: MP4 or MOV up to 50 MB. Audio: MP3 or WAV. Exact limits depend on the model.
Why AI Video Characters Drift and How to Fix It
When a face changes after the third or fourth clip, the cause is almost always the references or the prompt. Here is what to check.
| Problem | Why it happens | How to fix it |
|---|---|---|
| The face changes from clip to clip | The reference shows only one angle, or the lighting differs a lot between references | Use a turnaround sheet with front, side and back views, and reuse the same sheet and @ tags for every clip |
| Two people merge or an extra person appears | One image holds several people, or a turnaround sheet is read as several people | Use one image per character, and say that the sheet shows one person, for example: @image1 is a turnaround sheet of one girl |
| A reference is ignored | The prompt never mentions it | Call it by its tag and say what it does in the scene |
| The outfit or product changes | The prompt describes a different look than the reference | Remove clothing and color details from the prompt and let the reference show them |
| Logos or text look warped | The product image is small, blurry or shot at an angle | Use a sharp, straight-on product image |
| A photo of a real person gets blocked | Some models reject images that look like real people | Try another model such as Kling V4. Blocked videos refund your credits |
| Motion looks stiff | The prompt describes looks but not action | Describe the motion and camera movement, or add a reference clip |
Which AI Models Support Reference to Video?
Six AI models in the generator read references. They share one credit balance, so you can switch models when a result misses.
| Model | Reference uploads | Length | Aspect ratios | Max resolution | Best for |
|---|---|---|---|---|---|
| Kling V4 | Up to 10 images, 5 clips and 3 audio filesUp to 15 files in total | 3-30s | 16:9, 9:16, 1:1, 21:9 | 1080p | Characters across shots, with built-in sound |
| Seedance 2.5 | Up to 30 images, 10 clips and 10 audio files | 4-30s | 16:9, 9:16, 3:4, 1:1, 4:3, 21:9 | 1080p | The most references and the longest videos |
| Seedance 2.0 | Up to 9 images, 3 clips and 3 audio files | 4-15s | 16:9, 9:16, 3:4, 1:1, 4:3, 21:9 | 1080p | Mixing images, motion clips and audio in one shot |
| Seedance 2.0 Fast | Up to 9 images, 3 clips and 3 audio files | 4-15s | 16:9, 9:16, 3:4, 1:1, 4:3, 21:9 | 720p | Faster, lower-cost drafts |
| Seedance 2.0 Mini | Up to 9 images, 3 clips and 3 audio files | 4-15s | 16:9, 9:16, 3:4, 1:1, 4:3, 21:9 | 720p | The cheapest way to test a prompt |
| Grok Imagine Video | Up to 8 images | 3-15s | 16:9, 9:16, 3:4, 1:1, 4:3, 3:2, 2:3 | 720p | Quick results from images only |
Every model makes AI video with sound. Failed or blocked generations refund your credits automatically.
Reference to Video AI: Frequently Asked Questions
What is reference to video AI, and how does it work?
Reference to video AI creates a new video from reference images, clips or audio that show what your subjects look like and how they move. The AI analyzes each file for identity, style and motion, then generates a fresh shot from your prompt that keeps those details consistent.
How is reference to video different from image to video?
Image to video treats your picture as the first frame and animates forward from that exact composition. Reference to video uses your images as a visual reference only, so the subject can appear from different angles, in new lighting or in a whole new place. Choose image to video to animate one photo, and reference to video to put the same character or product into new scenes.
How many reference images can I upload?
Up to 30 with Seedance 2.5, 10 with Kling V4, 9 with Seedance 2.0 and 8 with Grok Imagine Video. Seedance and Kling V4 also take reference clips and audio files. Most shots only need 1–3 images, one per subject.
How do I refer to a specific reference in my prompt?
Each upload gets a tag in order: @image1, @image2, @video1, @audio1. Type @ in the prompt box to pick one, then say what it does, for example: @image1 wears the jacket from @image2. The same tags work on every model.
How do I keep a character consistent across clips?
Use a turnaround sheet that shows the character from the front, side and back, reuse the same sheet for every clip, and keep clothing details out of the prompt so they don't fight the reference. No AI video generator is perfect, so generate a couple of takes and keep the best one.
What is a character turnaround sheet, and how do I make one?
A turnaround sheet shows one character from the front, side and back in a single image. It gives the AI every angle of the face and outfit, so the character stays the same when they turn or move. You can make one with any AI image generator: describe your character and ask for a turnaround sheet with front, side and back views at the same size, on a plain background, with no text. Then upload it as one reference and tag it like any other image.
Can I use a YouTube video as a reference?
You can't paste a link, but you can upload a video file you have the right to use. Trim it to the few seconds of motion you need, save it as MP4 or MOV, and add it as a reference clip with Seedance or Kling V4.
Do the videos have sound? Can I use audio as a reference?
Yes. Every model here makes AI video with sound. With Seedance and Kling V4 you can also upload an MP3 or WAV as an audio reference, so the motion follows your music or voice line.
What length and aspect ratio are supported?
Videos run from 3 to 30 seconds depending on the model, at up to 1080p. You can pick 16:9 for YouTube, 9:16 for TikTok and Reels, 1:1 for feeds, and more.
Can I use a start and end frame with references?
Not in the same video. References and start and end frames are separate modes. If you need a fixed first frame or end frame, switch to the Frame to Video tab in the generator or use image to video.
Can I use reference to video AI for free?
Yes, you can start with free credits. New users get 10 free credits at sign-up. After that, each video costs credits based on the model, resolution and length. Videos made with free credits include a watermark, and paid videos don't.
Can I use the generated videos commercially?
Yes. You own the videos you create with PixExact and can use them for any lawful purpose, including ads, paid social and client work, on both free and paid plans. Just make sure you have the rights to the images and clips you upload as references.
Does reference to video replace video editing?
For short clips, often yes: you get a finished shot with motion and sound without filming or editing. For longer videos, generate each scene with the same references and join the clips in your usual editor.
More Tools
Explore more AI-powered tools to supercharge your creative workflow.
Start Creating with Reference to Video AI
Upload your character images, product images or reference clips and create videos that stay consistent from scene to scene. Free credits to start.