Animate old photos: turn a still family photo into a gentle memory video

What it does: takes one photo, optionally repairs visible damage and color loss on it, then produces a short video where the people in the photo show small, natural signs of life — a smile forming, an eye blink, a slight turn of the head. The restoration prompt is specific about the kind of damage it targets: scratches, dust, tears, stains and creases, plus faded or shifted color. It is also specific about what it must not do: it is told not to modernize the photo or change anyone's identity, only repair it.

What it does not do: it does not make the person talk, sing, dance or perform any scripted action — that is a different tool (AI Talking Avatar). It does not composite new people into the photo, change the background, or relocate the scene. It does not guarantee the restoration succeeds — if the AI restoration call errors out, the code quietly falls back to animating your original photo rather than failing the whole job, so a heavily damaged photo can sometimes come back animated but not visibly repaired. It also does not accept a custom animation prompt; you are choosing from four preset motion styles, not writing your own instruction.

This combination — real optional restoration, plus deliberately restrained animation, delivered only as a finished video — is what separates it from a plain "photo to video" tool and from a photo-editing tool that stops at a still image. If you only want the repaired photo itself with no motion, this is the wrong tool; there is no way to download the restored still on its own from this flow.

Stage 1 (restoration, optional, on by default) calls VdoBloom's "auto-image-edit" chain, an automatic pipeline that routes to Seedream 4.5 first, with a WAN model as a fallback if that call has trouble. It receives the uploaded photo plus a fixed restoration instruction and returns a repaired image URL, which becomes the input to stage 2. This step is billed and priced independently at a flat 7 credits, the same rate VdoBloom charges for any Seedream-4.5-tier automatic image edit elsewhere on the platform.

Stage 2 (animation, mandatory) sends the photo — restored or original, depending on whether stage 1 ran and succeeded — into an image-to-video generation call. The default model is ByteDance Seedance 2, but the tool exposes VdoBloom's normal video-model picker, so any image-to-video model enabled on the account (Kling, Runway, WAN and others, subject to VdoBloom's feature flags) can be used for this step instead of Seedance 2. Whichever model you pick, its own duration, resolution and aspect-ratio options and its own pricing apply to the animation stage.

Because the two stages are billed separately by two different systems, the total cost of a job is restoration cost (0 or 7 credits, depending on the toggle) plus whatever the chosen animation model charges for the duration and resolution you pick. There is no single flat price for the tool as a whole — the pricing table below shows the two components and the Seedance 2 defaults.

The obvious use is a memory video for family sharing: a grandparent's only surviving portrait, a wedding photo from decades ago, a photo of someone who has passed away. The restraint of the four motion presets (no dramatic movement, no added people) is specifically suited to this — it reads as respectful rather than novelty animation.

A second, more practical use is simply repairing a scanned or printed photo before you do anything else with it. Since the restoration step is a real AI image-edit call and the tool always shows you the intermediate consent and upload flow, some users run this specifically for the restoration effect, accepting the mandatory animation step as a side effect of getting the repair.

It is not built for marketing, product or creative animation — the fixed motion presets and the "no scene changes" instruction rule that out. For any use case that needs a written animation prompt or dramatic motion, VdoBloom's general Image to Video tool (with a free-text prompt) is the better fit.

Input is one photo per job; there is no cap on file size coded into this specific upload step (unlike some other VdoBloom upload flows that enforce a fixed MB limit). Output settings — duration, resolution, aspect ratio — are inherited from whichever image-to-video model you select for the animation stage; on the default Seedance 2 model that means 4 to 15 seconds, 480p through 4K, and several aspect ratios including the 9:16 default.

Restoration is a flat 7 credits regardless of photo size, charged only if you leave the restoration toggle on. Animation is priced per the selected model; the table below shows Seedance 2's tiered rates by resolution and length, since that is the model this tool defaults to.

Leave restoration on for any scanned print, faded photo, or photo with visible scratches or tears — it costs 7 credits and there is a fallback if it does not help, so there is little downside to trying it. Turn it off only for photos already in good digital condition, to save the 7 credits.

Pick the motion preset that matches the photo's framing: Gentle Memory and Coming Alive read naturally on solo portraits, while Family Moment is meant for group photos where more than one person should show a small movement.

Start with a shorter, lower-resolution animation (480p, 4 seconds) to see the result cheaply before spending more credits on a longer or higher-resolution version, since the two stages are billed independently and a failed high-resolution run is refunded but still costs time.

Upload the highest-quality scan or photograph you have — the restoration model is repairing an image, not inventing one, so more visible detail in the source gives it more to work with.