Image-to-video AI can turn a scanned family photo into a short, moving clip: a gentle smile, a turn of the head, wind in the trees behind. Done well, it feels like a memory coming back. Done badly, it looks uncanny. This guide shows how to get the first result.
What you need
- A scan or clear photo of the picture (at least 1,000 pixels on the short side if possible).
- An image-to-video AI tool, such as the TextVidAI generator.
- Permission from family members if the people in the photo are recognizable and the result will be shared publicly.
Step 1: Prepare the photo
- Scan, don't photograph at an angle. If you must use a phone, shoot straight on in daylight without glare.
- Crop to the subject. Remove borders and album edges.
- Restore before you animate. Fix scratches, tears and fading first. Animation amplifies damage, and a model may treat a scratch as an object.
- Decide on color. You can keep black-and-white or colorize first. Mixing both often looks odd.
Step 2: Write a prompt for natural motion
The biggest mistake is asking for too much movement. Old photos work best with small, believable motion. Use this structure: who is in the photo + one small action + camera + mood.
Prompt examples
- The woman in the photo slowly blinks and gives a soft smile, slight head tilt toward the camera. Very subtle motion, static camera, gentle film grain, nostalgic mood.
- The couple stands still while their clothes and hair move lightly in a breeze; leaves move in the background. Slow push-in, no other movement.
- The children laugh and look at each other, then back at the camera. Natural, subtle movement, warm afternoon light.
- Black-and-white portrait of a man in uniform. He breathes, blinks once and raises his chin slightly. Static camera, keep his face exactly as in the photo.
Step 3: Choose settings
- Length: 4–6 seconds is usually enough. Longer clips drift further from the original face.
- Aspect ratio: match the photo (most portraits look best in 9:16 or 3:4-style framing).
- Audio: for most memorial or family clips, choose ambient sound or none. Generated speech for real people can feel intrusive.
Step 4: Review and refine
- Face changes: reduce motion ("very subtle"), shorten the clip, and add "keep the face exactly as in the photo."
- Extra people or objects appear: crop tighter and describe only what is in the photo.
- Background warps: ask for a static camera.
- Too smooth or plastic: add "film grain, natural skin texture."
Using old photos respectfully
An animated photo is an interpretation, not a recording. Label it as AI-animated when you share it, don't put words in a real person's mouth, and ask relatives before posting animations of people they knew.
Related guides
- How to turn a photo into a video with AI
- First and last frame AI video
- Text to video vs image to video
FAQ
Can AI animate a black-and-white photo?
Yes. Describe it as a black-and-white photo in the prompt so the model keeps the look, or colorize it first.
How long should an animated old photo be?
Usually 4–6 seconds. Short clips stay closest to the original face.
Why does the face look different after animation?
The model has to invent views it cannot see. Reduce motion, keep the camera still and use a higher-quality scan.
Last updated: October 9, 2026.