Image to Video: Your Still, Finally Moving

Upload the frame you already approved and say what should move. Three models, one source image — cloth settles, liquid pours, the camera travels, and the shot still looks like yours.

Image to Video

Quicker renders for iteration and previews. Up to 4K and 20s clips at 24/25 FPS.

Start frame — JPG or PNG

End frame — optional

Max 1.25 MB per image, 2.5 MB total for start and end frames.

0/2000

0 credits available

Your clip appears here

Hit Generate to start the render on our cloud GPUs. The result plays back here as soon as it is ready.

The Frame Is Already Approved. Now Make It Move.

You have a product shot the client signed off on, a portrait that has been retouched, a key visual the whole campaign is built around. Image to video takes that exact frame as the opening shot and adds the motion you describe — the camera drifts, the fabric settles, the liquid pours — while subject, palette, and composition stay recognisably yours. Choose LTX 2.5 for 4K and native sound, Seedance 2.0 for social ratios, or Kling 3.0 when you want to set the end frame too.

Demo
Source image uploaded for the image to video demo

Input · Source Image

One still, uploaded as the first frame. Prompt describes only the motion.

Generated

How to Turn an Image Into Video

1

Choose a video model

Select LTX 2.5, Seedance 2.0, or Kling 3.0. Resolution, duration, and audio options update to what that model supports for image input.

2

Upload the image and describe the motion

Drop in the still you want to animate — it becomes the first frame. Then write a short prompt about what moves: camera, subject, atmosphere. You are directing motion, not re-describing the picture.

3

Generate, compare, download

Preview the clip in the browser. If the motion is not right, adjust the prompt or switch models; the source image carries over unchanged. Download when it matches.

Why Start From an Image Instead of a Prompt

With text to video, the first frame is a lottery: you describe the scene and hope the model lands on the right look. With image to video, that decision is already made. You bring an approved product photo, a portrait that has been retouched, a concept frame the client signed off on — and spend your generations on motion, pacing, and atmosphere rather than on re-rolling the composition. For teams working with existing brand assets that is the difference between a usable clip on the second try and a usable clip on the tenth.

Why Teams Animate Their Stills Here

The first frame is a decision, not a roll of the dice

Starting from an approved image removes the biggest source of wasted generations. You already know what the shot looks like; the only thing left to get right is the motion.

Three models, same source image

Because the uploaded still carries over when you switch models, comparing LTX 2.5, Seedance 2.0, and Kling 3.0 is apples to apples. Same frame, same prompt, different interpretation of the motion.

Controlled motion for commercial assets

For ads, landing pages, and social content, a subtle orbit or a slow reveal is usually worth more than dramatic animation. The prompt gives you that control instead of leaving it to chance.

Picking a Model for Image to Video

All three models accept an image as the first frame, but they behave differently. LTX 2.5 is the pick when the output needs to be large — up to 4K — or when you want ambient sound rendered in the same pass. Seedance 2.0 is quick and handles vertical and square formats well, which suits social cut-downs of a product shot. Kling 3.0 gives you longer 1080p takes, useful when one still needs to carry a slow reveal or orbit. Because the source image stays the same, switching models is a fair comparison: same frame, same prompt, different motion. If you have no image yet, start with text to video; if the clip needs to follow a soundtrack, use audio to video.

FAQ

What does image to video actually do?+
It takes a still image as the opening frame and generates a video clip from it, guided by a short prompt describing the motion, camera movement, and mood. The image you upload is the image the clip starts on.
Which video models support image to video here?+
LTX 2.5, Seedance 2.0, and Kling 3.0 all accept an uploaded image as the first frame. You can switch models from the selector and keep the same image and prompt.
What kinds of images work best?+
Clean, well-lit images with a clear subject: product photos, portraits, concept art, illustrations, campaign visuals. Busy or low-resolution images give the model less to anchor on.
How much of my original image is preserved?+
The uploaded image is used as the first frame, so composition, subject, and palette start exactly as you provided them. How far the scene drifts over the clip depends on the model and how aggressive your motion prompt is.
How do I write a good image to video prompt?+
Describe motion, not the picture — the model can already see the image. Say how the camera moves, what the subject does, and what the atmosphere should feel like. Keep it to a few sentences.
Is image to video better than text to video?+
Image to video is better when you already have a visual you want to keep. Text to video is better when you are inventing the scene from scratch and have no source frame.
Can the output include sound?+
Yes, on models that support audio generation, such as LTX 2.5 and Seedance 2.0. Enable audio in the options and the sound is rendered together with the video.

Use the Image You Already Trust and Turn It Into Motion

Upload the product shot, portrait, or concept frame you already have, write one line about what should move, and watch it play back with sound. If the motion is not right, change the prompt or the model — the frame you uploaded stays exactly as it is.

See Plans and Credits