Free tool · no sign-up
Photo readiness checker
Upload a photo to see how ready it is for AI video trends: resolution, framing and a few estimated quality signals. Scores are guidance — they do not guarantee a successful generation.
JPG / PNG / WebP up to 15 MB. No sign-up needed — checks are rate-limited per visitor, and your photo is not stored.
File format, resolution and framing are measured from your image. Face, sharpness and lighting are marked estimated — they are guidance, not a real face detection.
What the checker looks at
The readiness check runs the same five tests the generator applies, so a green result here is a reliable preview of how your photo will behave in a real generation. Each check maps to a specific way the AI model reads your image — and each failure mode has a fix you can apply in any photo editor.
1 · File format
The model accepts JPG, PNG and WebP. Screenshots, heavily compressed exports and exotic formats lose detail before generation even starts. Re-export at full quality if this check fails.
2 · Resolution
The face needs enough pixels for the model to keep it consistent through motion. As a rule of thumb the head should be at least 200 pixels tall — photos straight from a modern phone always clear this, screenshots from social apps often do not.
3 · Framing
A chest-up or waist-up crop gives the model room to animate gestures. A tight passport-style headshot leaves nothing for the shoulders and hands to do, and a full-body shot makes the face too small to hold.
4 · Face visibility
The face should be front-facing and unobstructed. Sunglasses, heavy shadows across the eyes, and profiles force the model to invent features it cannot see — which is where likeness drift starts.
5 · Lighting
Even, directional light reads best. Mixed lighting (warm indoors plus cool window light) and hard shadows make the model guess at skin tones, which shows up as flicker in the generated clip.
Common failures and how to fix them
"Face too small" — crop tighter around the subject before uploading. The fix takes ten seconds and improves every downstream check.
"Low resolution" — find the original photo instead of a copy from a messaging app, which recompresses images aggressively. If the original is genuinely small, generate at 768P rather than 2K.
"Harsh lighting" — shoot again near a window, or use the editor's auto-exposure. Do not rely on filters — most of them flatten the shadow detail the model needs.
"Group photo" — crop to a single person. Every template assumes one clear subject per photo (two photos for two-person trends), and extra faces confuse the framing.
After the check passes
A passing photo is ready to generate — pick a trend, choose a model and length, and the live cost counter shows the exact price before you confirm. If a result still disappoints with a passing photo, change the scene or the length before blaming the source: a better input fixes more than a bigger budget, but scene choice fixes more than either. Run the check again whenever you switch photos — it is free, and it takes seconds.
When to run the checker again
The check is not a one-time ritual — it is a habit for every new photo, and it is worth re-running in a few specific situations. After editing a photo in any app: edits change compression, color and framing, and a photo that passed as an original can fail after an aggressive filter or a bad crop. After switching devices: cameras differ more than people expect, and a phone that produced perfect inputs may produce soft ones. Before re-rendering a keeper: the photo that generated a good 4-second test will generate a good 15-second version — but only if it is the exact same file, not a re-saved copy.
The checker also cannot tell you things no automated tool can. Whether the person in the photo consents to being in your video, whether a scene choice fits the memory you are recreating, whether a trend is the right format for what you want to say — those stay human decisions, and the content policy is there for the lines that must not be crossed.
Myths worth retiring
"AI fixes bad photos." It can stylize them, but likeness consistency is bounded by input detail — a soft face stays soft, just animated. The check exists so you learn this before spending.
"Higher resolution always means better output." Up past the point where the face has detail, extra pixels add cost without adding likeness. 768P on a good 800-pixel-tall face beats 2K on a 200-pixel one every time.
"Professional cameras are required." Lighting and framing beat hardware. Recent phones in decent light pass every check; studio cameras in bad light fail them.
A quick checklist before you upload
Do
Use the original file, straight from the camera or full-quality export · face front-facing and evenly lit · chest-up framing · run the checker once more after any edit · keep the passed file as your generation master.
Don't
Don't screenshot from social apps · don't upload screenshots of screenshots · don't use beauty filters that repaint skin texture · don't crop so tight the head touches the frame · don't fix one warning and ignore the other three.
Privacy of the check
The check runs against the photo you upload, over an encrypted connection, and the result is stored only as a status on your own asset record — the photo itself is not shared, published, or used for anything beyond the check you asked for. You can run it as many times as you like on as many photos as you like; nothing is charged and nothing counts against your generation limits.