How to Turn a Photo into a Video
Updated Sep 2, 2026 · 8 min read
What image-to-video actually does
Image-to-video takes a single still frame and generates the frames that would follow it, guided by a short text description of the motion you want. Unlike a slideshow effect (Ken Burns pan and zoom), the model understands what is in the picture: a portrait can blink and turn slightly, water can ripple, clouds can drift, a product can rotate, and a camera can appear to dolly forward through the scene.
The Image to Video tool wraps five models with different price and quality points:
- Wan 2.2 (economy), 8 credits: 5 seconds, 480p to 720p. Excellent value, good for social posts and drafts.
- LTX-2 (fast), 6 credits: fastest generation, ideal for testing prompts before spending on a premium model.
- Seedance 1 Pro, 12 credits: strong prompt adherence and multi-shot storytelling.
- Kling 2.5 Turbo Pro, 15 credits: cinematic camera motion at 1080p, the best all-rounder for ads.
- Veo 3.1, 60 credits: top fidelity with native audio, 8 seconds. Reserve it for the final render.
Free accounts can use Wan and LTX; Kling and Seedance unlock on Pro, and Veo on Business. The right workflow is to iterate cheaply and render once at the top.
Step 1: Prepare the still image
The output can only be as good as the input frame, because the first frame of the video is essentially your photo.
- Resolution. Upload at least 1280 px on the long side for 720p output and 1920 px for 1080p. If your photo is smaller, run it through the Image Upscaler first; 2x costs 1 credit and it is far cheaper than a disappointing 15-credit video.
- Aspect ratio. Crop the image to the ratio you will publish in: 9:16 for TikTok, Reels and Shorts; 16:9 for YouTube and web; 1:1 for feed posts. The tool offers all three. Cropping beforehand avoids the model inventing content at the edges to fill a new ratio.
- Composition. Leave some room around the subject. If the subject touches the frame edge, any camera move will push them out of frame.
- Clean image. Remove text overlays and watermarks from your own images first; the model will try to animate them and the result looks broken.
- One clear subject. Busy group shots animate less predictably than a single person or product.
Step 2: Write a motion prompt that works
The prompt describes motion, not the image. The model already sees the image; telling it "a woman in a red dress" wastes tokens. Tell it what changes over time.
A reliable structure is: subject motion + camera motion + atmosphere. For example:
- She slowly turns her head toward the camera and smiles softly. Camera pushes in gently. Warm afternoon light, hair moving in a light breeze.
- The sneaker rotates 90 degrees on its axis, studio lighting, subtle reflections on the floor, camera static.
- Clouds drift right to left, water ripples, slow drone pull-back revealing the coastline.
Rules that consistently help:
- One or two motions, not five. Overloaded prompts produce jittery, confused results.
- Name the camera move explicitly: static, slow push in, pull back, pan left, orbit, handheld. "Static camera" is underrated; it makes subject motion look more natural.
- Use speed words: slowly, gently, subtle. Models tend to over-animate by default.
- Say what should not move if it matters: the logo remains sharp and unchanged.
- Avoid impossible physics for the still: a photo of a closed door will not convincingly show what is behind it.
Test the prompt on LTX-2 (6 credits) or Wan (8 credits) first. Once the motion reads correctly, switch to Kling or Veo for the final.
Step 3: Choose duration, resolution and generate
Set duration to 5 seconds for most social clips; 8 or 10 seconds only when the motion genuinely needs it, because longer clips have more opportunity to drift from the source image. Set resolution to 720p for drafts and 1080p for delivery (Kling and Veo). Pick the aspect ratio matching your crop.
Click Generate. Typical wait times: LTX about 20 to 40 seconds, Wan around a minute, Kling one to three minutes, Veo two to four minutes. You can queue several variations and leave the tab; finished jobs appear in your history.
Review the clip frame by frame at the start and end. The most common defects are:
- Face drift: the person looks slightly different by frame 100. Shorter duration and "subtle" wording reduce it.
- Morphing hands and text: hands that gain fingers, letters that shift. Keep hands still in the prompt, or crop them out.
- Looping seam: if you plan to loop the clip, ask for motion that returns to the start, for example gentle sway, returning to the original pose.
Generate two or three seeds of the same prompt; picking the best of three is standard practice, not a sign of failure.
Step 4: Export and use it
Download the MP4 (H.264, with AAC audio when using Veo). No watermark is applied on paid plans. For platform uploads:
- TikTok / Reels / Shorts: 9:16, 1080 x 1920, 5 to 10 seconds; add music and captions in the platform editor.
- Product ads: 1:1 or 4:5 for feed, 9:16 for stories. A 5-second rotating product with a static camera performs reliably.
- Website hero: 16:9, export at 720p to keep file size down, loop it muted.
- Talking presenters: if you need speech, use the Talking Photo tool instead; image-to-video does not lip-sync.
If you want to extend a good clip, take its last frame as a new source image and generate the next 5 seconds with a continuing prompt. Stitching two or three of these gives a 15-second sequence at Wan prices.
For text-only starts with no photo at all, the AI Video Generator uses the same models from a prompt alone.
Frequently asked questions
How long can the video be?+
Five, eight or ten seconds per generation depending on the model. Chain clips by using the last frame as the next source image for longer sequences.
Which model should I start with?+
LTX-2 at 6 credits for prompt testing, then Kling 2.5 Turbo Pro at 15 credits for a 1080p final. Use Veo 3.1 only when you need native audio or maximum fidelity.
Does the video have sound?+
Veo 3.1 generates native audio. The other models produce silent clips; add music or voice-over in your editor.
Can I use the videos commercially?+
Yes on Pro and Business plans, which include a commercial licence for generated outputs. You must own or have rights to the source photo.