Dance from a video referenceSeedance 2.0
Inputs: a character photo and a short dance video. The model transfers the motion from the reference to the character, so the movement is defined visually instead of described word by word.
Animate a photo and turn it into a short AI video
Upload a product image, portrait, or illustration and describe the motion: camera, light, gesture, or action. Seedance, Kling, Veo, Wan, and HappyHorse turn the still frame into video while preserving the main subject.
Animate an imageNo card required. Upload a photo and test the first motion variant.
Image to video is the right mode when you already have a frame and need to bring it to life: a portrait, product, character, interior, or ad visual. You are not inventing the scene from scratch; you start from the composition you already trust.
The result can stay in the same Givon AI project: save it to the library, use it as a scene, add voiceover and captions, or include it in a longer video without moving files between tools.
If your task is different
This page covers the image-to-video step. Nearby workflows handle image creation, video generation from text, and final editing.
Example: a photo follows motion from a video reference
A real Givon AI generation: one source photo and a short video reference produce a moving version of the character.
How to write the prompt
A prompt should describe content: subject, action, camera, light, and style. Duration, aspect ratio, and resolution are generation settings, not prompt text.
Weak
Animate this photo beautifully, vertical, 5 seconds.
Better
The camera slowly moves closer, soft light passes over the product surface, the background gently blurs, the object stays stable, commercial style.
Models for image animation
Several catalog models accept an image as the starting frame, and each has its own motion style and controls.
Seedance 2.0
4-15s · native audio · start and end frames · up to 9 references
Start and end frames, motion from video references, and native sound in the same generation.
Kling 3.0
3-15s · native audio · start and end frames · up to 2 references
Careful motion and camera control for products and portraits.
Veo 3.1
8s · native audio · start and end frames · up to 3 references
Cinematic eight-second takes from a single image.
HappyHorse 1.0
3-15s · native audio · up to 9 references
A strong option for faces, emotions, and portrait animation.
Wan 2.7 Video
2-15s · start and end frames
First-to-last-frame mode when the motion must land on a specific ending.
The catalog keeps evolving, and available modes plus token costs depend on the current publication. The full list is inside the editor.
How it works
- 01
Prepare the image
Choose a clear image with one main subject that does not disappear into the background or get blocked by details.
- 02
Describe the motion
Define what happens: camera movement, gesture, turn, new detail, or lighting change. Keep one main action.
- 03
Generate variants
Compare rhythm, subject stability, and composition. The strongest result is rarely the first one.
- 04
Save the strong take
Move it to the library if it can help the next scene or a series of videos.
- 05
Add context
Connect the next scene, voiceover, and captions, or keep the clip as a standalone short video.
- 06
Review before publishing
Check faces, hands, text, and product shape because small details may shift when the image starts moving.
Best use cases
Product ads
Animate a static product card with camera and light motion for social promos.
AI characters
Turn a portrait or character image into a short video take for recurring content.
Content without a shoot
Use existing photos and references when there is no time or budget for filming.
One strong frame
When the composition is already right, animation is the shortest path from image to video.
How to choose a source image
The best source images have a clear main subject, enough detail, and a clean relationship with the background. If a specific person, product, or logo matters, the source must be sharp.
If the frame is too complex, split the job into steps: animate the main subject first, then build the next scene separately. This keeps each motion under control.
Why Givon AI
Built on an existing image
You start from a frame that already contains the product, character, background, and composition.
Main subject control
The prompt can explicitly say that the subject stays stable while only the camera, light, or background moves.
Different models for different motion
Seedance, Kling, Veo, Wan, and HappyHorse animate images differently. If one model misses the movement, try the same task in another.
Materials stay in the project
The source and result are saved to the library, ready for another scene or a repeated motion.
What to keep in mind
Quality depends on the task, source materials, and number of attempts. Here is what to account for before publishing.
Small details may change
Texture, background, hands, and text can be interpreted by the model, especially with strong motion.
One image, one main motion
A clear subject and one action give the most stable result. Split multi-event scenes into steps.
Product photos need subject stability
If shape, packaging, or logo matters, ask the model to keep the object stable and review the output before publishing.
Use care with portrait expression
For faces, start with light camera movement or a small gesture. Strong expression changes can alter features.
Source image rights still matter
If the photo includes a person, brand, logo, or someone else's work, check permissions before advertising or publishing.
FAQ
Can I animate any photo?
Technically, almost any image can be tried, but clear images with one main subject work best. Blurry or overloaded frames are less stable.
Can I make a video from a product photo?
Yes. A product photo can become the starting frame while you describe camera and light motion. Review product shape and packaging in the result.
Which models animate images?
Seedance, Kling, Veo, Wan, HappyHorse, and other catalog models accept an image as a starting frame. Some also support an ending frame.
Why can the result differ from the source?
The model invents motion and may interpret the background, hair, hands, or texture. Review important details before publishing.
Can I add voiceover and captions later?
Yes. The animated frame remains a scene in the Givon AI project, where you can add voiceover, captions, more scenes, and export.
What if the first result is not good enough?
Simplify the motion, clarify the prompt, or change the model, then generate a few more takes. Short test loops are normal for video.
How long is the animated video?
It depends on the model: Seedance 2.0 supports 4 to 15 seconds, Veo 3.1 creates 8-second takes, and Wan 2.7 can go up to 15 seconds. Longer videos are assembled from several scenes.
Is it free to try?
Registration does not require a card. Starter tokens cover the first generations; after that you can use a plan or buy more tokens. Unused tokens do not expire.
Create your first result in Givon AI
No card required. Upload a photo and test the first motion variant. After the first result, save the asset and keep working in the same project.
Animate an image