Want to try the AI version right now? A browser tool that animates a still image turns a JPEG, PNG or WebP into a short clip from a one-line description. It makes short clips (2 to 5 seconds) at low resolution, with no watermark. It runs on the servers of a sister site from the same publisher, so the image is uploaded; the rest of this page explains what to expect and the offline alternatives.
Three ways to turn an image into a video
| Method | What moves | Cost and effort | Best for |
|---|---|---|---|
| AI image to video | Things inside the picture: water, hair, smoke, a head turning, plus camera motion | Free tiers exist; short clips; image is uploaded to the service | Social clips, animated portraits, product teasers, B-roll |
| FFmpeg pan and zoom (“Ken Burns”) | Only the virtual camera; the picture itself is unchanged | Free, offline, exact, needs a terminal | Archive photos, documentaries, anything that must stay faithful |
| Slideshow or editing app | Camera motion, transitions, titles, music | Free apps exist; no terminal | Several photos with music, birthday and memorial videos |
A quick test: if the picture must not change at all (an archive photo, a legal exhibit, a product label), use FFmpeg or a slideshow app. If you want the scene itself to come alive, use AI and accept that the model is inventing the in-between frames.
Method 1: AI image to video, step by step
- Choose a photo you have the right to use. Your own photo, a licensed image, or an image you generated. One clear subject on a fairly calm background animates better than a busy group shot.
- Crop to the target shape first. Decide where the clip will play: 9:16 for Reels, Shorts and TikTok, 16:9 for YouTube and slides, 1:1 for some feeds. Most generators also ask for an aspect ratio; cropping beforehand means you decide what stays in frame, not the model.
- Use a format the generator accepts. JPEG, PNG and WebP are the safe set. If your file is AVIF, GIF or BMP, convert it with the in-browser image converter on this site; it runs locally and does not upload the file. iPhone HEIC photos are easiest to export as JPEG from the Photos app (or set the camera to “Most Compatible”), because most browsers outside Safari cannot open HEIC. Very large images do not help, because free AI video models output roughly 480p, but a soft or blocky image will look worse once it moves.
- Describe the motion, not the picture. The model can already see the photo. Tell it what should move and how the camera behaves, in one sentence: “The waves roll in slowly, foam spreading on the sand, the camera holds still.” One main action per clip.
- Generate several takes. Upload the image, pick a model, the aspect ratio and the length, and run it. You can animate the photo with an AI video generator in the browser: it offers four open models (LTX 2.5, LTX 2.3, Wan 2.2 14B and Wan 2.2 5B), clips of 2 to 5 seconds at 480p to 512p depending on the model, and MP4 output without a watermark, on a free daily allowance. Not every model accepts an input image; the upload field appears when the chosen one does. Run the same prompt three or four times and keep the best.
- Check it frame by frame. Scrub slowly through the clip in a player that can step frames (VLC, mpv, or your editor). Look at hands, eyes, teeth, edges, logos and any text. Warped lettering and melting fingers are the usual failures.
- Prepare the file for its destination. Most platforms accept the MP4 as it is. If a phone or an older device refuses it, re-encode with the settings in convert video for mobile. If you are joining several clips, do that in an editor and export once.
Motion prompts that work
| Image | Prompt | Why it works |
|---|---|---|
| Portrait | “She blinks and smiles slightly, a light breeze moves a few strands of hair, slow push-in.” | Small motion keeps the face recognisable |
| Landscape | “Clouds drift left to right, grass sways, slow aerial drift forward, late afternoon light.” | Natural elements are what the models do best |
| Product on a table | “The camera orbits slowly around the bottle, soft studio light, background stays still.” | Camera motion, not object motion, keeps the label intact |
| Illustration or poster | “Gentle parallax: foreground leaves move slightly, background stays, subtle glow pulses.” | Layered motion suits flat artwork |
| City at night | “Car light trails pass below, neon signs flicker, the camera tilts up slowly.” | Light and traffic are easy, readable motion |
Words like “epic”, “beautiful” or “cinematic” on their own say nothing about movement. Speed words help (“slowly”, “time-lapse”, “slow motion”), and so does naming the camera move (“static camera”, “push-in”, “orbit”, “tilt up”). If a prompt in another language is ignored, try the same sentence in English; many models were trained mostly on English captions.
Method 2: a free pan-and-zoom with FFmpeg
If the picture must stay exactly as it is, you do not need AI. FFmpeg can turn a still into an MP4 offline, and the result is predictable. The first command makes a 5-second, 1080p clip of the still image, letterboxed if the shape does not match:
ffmpeg -loop 1 -i photo.jpg -t 5 -vf "scale=1920:1080:force_original_aspect_ratio=decrease,pad=1920:1080:(ow-iw)/2:(oh-ih)/2,format=yuv420p" -r 30 -c:v libx264 -movflags +faststart still.mp4
The second adds a slow zoom toward the centre (the classic Ken Burns move), filling a 16:9 frame. The image is first scaled up to 4K so the zoom stays smooth, then cropped:
ffmpeg -i photo.jpg -vf "scale=3840:2160:force_original_aspect_ratio=increase,crop=3840:2160,zoompan=z='min(zoom+0.0015,1.3)':d=150:x='iw/2-(iw/zoom/2)':y='ih/2-(ih/zoom/2)':s=1920x1080:fps=30,format=yuv420p" -c:v libx264 -movflags +faststart kenburns.mp4
d=150 at 30 fps gives 5 seconds; raise it for a longer clip. zoom+0.0015 controls speed and 1.3 the final zoom. For a vertical clip, swap the sizes to 2160:3840 and 1080x1920. Both commands were tested with FFmpeg on a 3000 × 2000 JPEG.
Method 3: a slideshow or editing app
For several photos with music and titles, an editor is quicker than either of the above. Free options include the Photos app on Windows and macOS, Clipchamp on Windows, iMovie on Apple devices, CapCut, Shotcut and the free version of DaVinci Resolve. All of them can add a pan-and-zoom per photo, transitions, text and a soundtrack, and export an MP4. They do not generate motion inside the picture; for that, render AI clips first and drop them onto the same timeline.
AI image-to-video tools compared
Prices and free allowances change often, so this table describes what each option is suited to rather than what it costs today. Check the vendor’s page before you pay.
| Option | Account | Good at | Watch out for |
|---|---|---|---|
| Upsampler free video generator (sister site) | No | Trying image to video in seconds, no watermark, open models | 2–5 s clips at 480p–512p; longer clips and up to 1080p are paid |
| Runway, Kling, Luma Dream Machine, Pika, Hailuo | Yes | Strong commercial models, camera controls, extending clips, higher resolution | Free credits are limited; many features sit behind a subscription |
| Google Veo (in Gemini and Flow), OpenAI Sora | Yes | Very realistic motion; some generate sound | Availability depends on country and plan |
| Open models on your own GPU (Wan 2.2, LTX via ComfyUI) | No | Files stay on your machine, no quotas | Needs a capable graphics card and some setup |
| FFmpeg / slideshow apps | No | Faithful, offline, any length | Camera motion only, nothing in the picture moves |
Common mistakes
- Asking for too much motion. A full turn of the head or a walk across the frame from a single photo usually distorts the subject. Subtle motion looks more real.
- Several actions in one prompt. “He waves, laughs and sits down” produces muddy, confused motion. One action per clip, then edit clips together.
- Letting the generator crop. If you upload a 4:3 photo and ask for 9:16, the model decides what to keep. Crop first.
- Animating text and logos. Lettering tends to wobble or mutate. Keep the camera moving and the label still, or add text afterwards in an editor.
- Upscaling the photo instead of the clip. Free models output around 480p regardless of input size. If you need a sharper result, enlarge the finished video, not the input still.
- Re-encoding repeatedly. Every lossy encode costs quality. Edit, then export once.
Rights, people and labelling
This site’s rule for video applies here too: start from material you are allowed to use. Animating a stranger’s photo or a copyrighted artwork does not become lawful because a model generated the frames. If real people are recognisable, get their consent before you publish, and do not use AI motion to make anyone appear to say or do something they did not. Most generators prohibit impersonating real people in their terms.
Label realistic AI video when you publish it. YouTube, Instagram and TikTok all have disclosure settings for realistic synthetic content, and in the EU the AI Act’s transparency rules for deepfakes apply from August 2026. A short “animated with AI” in the caption is enough for most uses.
Related on this site: convert video for a phone, video for iPod and iPhone, and the legal options for YouTube video if your source is a clip rather than a photo.
FAQ
Can I turn an image into a video for free?
Yes. Without AI, FFmpeg or any slideshow app will turn a still into an MP4 with a slow zoom or pan, free. With AI, some browser generators animate a photo without an account or a watermark, on a daily free allowance; the clips are short (a few seconds) and low resolution, and longer or higher-resolution clips usually need a paid plan.
What is the difference between a slideshow and AI image to video?
A slideshow moves a virtual camera over a picture that does not change. AI image to video generates new frames, so things inside the picture move: hair, water, clouds, a turning head. Use the slideshow when the picture must stay exactly as it is; use AI when you want motion inside the scene.
How long can an AI image-to-video clip be?
A single generated clip is usually between 2 and 10 seconds. Free tiers tend to sit at the short end. For anything longer, generate several clips and join them in an editor.
What image format should I upload?
JPEG, PNG or WebP are accepted almost everywhere. Convert HEIC, AVIF, GIF, BMP or TIFF first, crop to the aspect ratio you want, and avoid tiny or heavily compressed images, because artefacts get animated too.
Why does the face change when I animate a portrait?
Every new frame is a prediction. The more motion you ask for, the further the model drifts from the original face. Ask for subtle motion (a blink, a small smile, a slow push-in) and pick the best of several takes.
Can I use the clips commercially?
It depends on the tool and its model licences, and on your rights to the photo. Some open models claim no rights over the output, but they cannot give you rights to an image or a person you did not have permission to use. Read the terms of the tool you use.