Upload an image, describe the motion, and turn still visuals into short AI videos with clear model and parameter controls.
Upload a still, describe the movement, and the model animates it faithfully — the use case communities love most, from reviving old family photos to turning product shots into ads.
The most beloved image-to-video use case on the internet: a vintage family photo gains gentle, natural motion — a smile, a breeze, a turn of the head. The original stays recognizable; the moment becomes a memory you can watch.
One hero shot becomes a moving clip — dust, fabric ripple, rotation, splash — ready for feeds and product pages. E-commerce teams turn an existing photo library into a video library without a single shoot.
Hair and clothing move with physics viewers accept; expressions stay true to the person. Kling 3.0 can even transfer a full motion reference — a dance, a gesture — onto your character while preserving identity.
Waterfalls flow, clouds drift, rain falls — landscapes and interiors become ambient loops for backgrounds, screens, and mood pieces, with native audio if you want the scene to sound like it looks.
Each of these began as a still image. The motion prompt that animated it is printed underneath.
Subtle living portrait: she blinks, exhales, and turns her head a few degrees toward the window light. Hair strands drift, golden dust motes float, the landscape outside slides past. Keep her exact face, freckles, and hair unchanged.
Image-to-video from a Seedream 5 Pro still: Subtle cinemagraph: the woman blinks softly and turns her head slightly toward the window light, hair moving in a gentle breeze, train scenery drifting past outside, film grain mood preserved.
Image-to-video from a Seedream 5 Pro still: Slow elegant orbital camera move around the perfume bottle, light glinting across the amber glass, subtle sparkle particles drifting, luxurious mood.
A seamless loop: the camera drifts a few centimetres closer and then settles back exactly where it started, her expression softening and returning. No cuts, no new elements.
Place the woman from @image1 walking slowly up the stone steps toward the red torii gate seen in @image2, early morning mist, lanterns glowing, her face and hair exactly as in @image1. Handheld documentary framing, birdsong and distant temple bell.
Engines built to animate a still without redrawing the person in it.
| Input | One image (JPG or PNG) as the first frame, plus an optional motion prompt. |
|---|---|
| Best-suited models | Seedance 2.5 for long clips with audio, Seedance 2.0 family for everyday work, Kling and Hailuo for stylised motion. |
| Duration | 4 to 30 seconds depending on the model; short clips hold the source frame most faithfully. |
| Resolution | 480p and 720p across the range; 1080p and 4K on Seedance 2.0 standard. |
| Audio | Native synced audio on supported engines, marked before you spend credits. |
| Cost | Per second of output, from 10 credits/second for reference-driven generation at 480p. |
| From your history | Any image you generated on the site can be sent straight to video without re-uploading. |
Upload a still, describe the motion, generate the clip.
A photo, an illustration, or something you generated a minute ago. The image to video generator uses it as the first frame.
Say what should move and how the camera behaves. Restraint reads better than spectacle — small, believable motion holds up where dramatic action tends to warp.
Download the clip, extend it into a longer sequence, or add synced audio depending on the engine.
The model already has your picture. The prompt only needs to describe what moves — and that is where most first attempts go wrong.
The one element that should animate. Everything you do not mention should stay put.
her hair and the hem of her coat lift in the wind
Speed and quality of the motion. “Slowly” and “gently” are doing real work here.
slowly, a gentle drift
Whether the frame moves. “Locked-off shot” is a valid and often better answer.
slow push-in, subtle
What must not change — usually the face and the setting.
keep her face and the background unchanged
make this photo move
gentle wind moves her hair and scarf, she blinks once and smiles slightly, camera holds still, background stays unchanged
The vague version leaves the model to decide what animates, which usually means everything animates — including the face, which is where warping shows first. The specific version names three small movements and locks everything else down.
make the car drive fast and the camera fly around it
the car rolls forward slowly, wheels turning, camera drifts right at a steady speed, reflections shifting across the paint
Fast subject motion combined with a moving camera is the hardest thing to ask for and the most likely to warp. Slowing both down keeps the geometry intact and reads more cinematic anyway.
Animating a photo is one step. Generation, editing, and avatars share the same account.
Which engines animate a still most faithfully, and what each costs.
Per-second pricing compared across the image to video engines.
Reference-driven video, resolutions, and pricing in detail.
The Pro engine’s take on animating a still.
Generating the still you are about to animate.
A fast option when you need the clip in a hurry.
Image to video is billed per second of output, so a short animation stays inexpensive. Free accounts get 80 credits a day.
Perfect for getting started
Billed annually · $240/year
Billed annually · $300/year
Upload a photo, describe the motion you want, and the model generates a video that starts from — and stays faithful to — your image, adding movement, camera work, and optionally sound.
Yes — fidelity to the source image is the core requirement, and the lineup (Seedance, Veo, Kling) is selected for exactly that. Faces and products remain recognizable.
Yes — reviving vintage photos is one of the most popular workflows. For badly damaged photos, restore them first with the AI image editor, then animate.
Yes — describe the movement in the prompt, or use Kling 3.0 Motion Control to transfer the exact motion from a reference video onto your character.
Per second of output depending on model, resolution, and audio, always shown before generating. Failed generations refund automatically.
Yes — tell Ava "animate this photo" in chat with the image attached, and it plans an image-to-video generation with the right model and settings.
MP4 video at the resolution you selected, stored in your history and downloadable from our CDN.
Sharp stills with a clear subject and some visual depth. Very low-resolution files, heavy motion blur, or dense crowds give the model too little to work with, and the resulting motion tends to smear.
The engines here are built to hold the first frame, so a face usually stays recognisable through short clips. Drift increases with duration and with how much motion you ask for — a 5-second subtle movement holds far better than 15 seconds of action.
Yes, and it is one click — any result in your history can be sent straight into image to video without re-uploading. Prompt and model of the original image stay recorded alongside the clip.
One-click tools for everything around generation — polish, restyle, and share.
