Best photos for AI trend videos: photo and consent checklist

The best photo for an AI trend video is a recent, sharp, solo photo of one adult, with the whole face visible, even light and no heavy filters, that you have the right to use. On Framecomet every video effect takes one photo per person, so a two-person scene needs two separate solo photos rather than one group shot. Everyone pictured must be 18 or older and must have agreed to be included. This checklist covers both halves of the job: photos that work, and the permission behind them.

Start with consent, age and rights

Before you think about lighting, check that you may use the photo at all.

• Everyone pictured is an adult (18+).
• Everyone pictured has agreed to appear in an AI-generated clip, and ideally knows where you plan to post it.
• The photo is yours, or you have permission from the person who took it and the people in it.
• No celebrities, public figures or minors, and no photos of strangers taken from the internet. Being publicly visible is not the same as permission.

When you add photos, the page asks you to confirm these points. Automated checks also screen each photo and estimate age, but they are a safety net, not a substitute for asking. Repeated blocks have consequences: three safety blocks within 24 hours pause an account's uploads and generations for 24 hours. The acceptable use policy lists everything that is not allowed.

How many photos each effect needs

Count the people in the scene and prepare one photo for each of them.

• Two photos, left and right: Raindance and Hotel Lobby.
• One photo per person, for the group size you choose: Gas Station Dance. Check its page for which group sizes are open.
• One photo of one adult: Ceiling Fan Transition, Car Driving, Cartoon Clone Dance, AI Crying, Storm and Cinema Breakout.
• One photo of one to six adults, delivered as a still image: 1980s Photo.

The AI video effects hub lists every scene, and each effect page shows how many photos it needs and its current status.

What makes a photo work

A generated scene can only use what is visible in your photo. These qualities matter most:

• Sharp focus on the face. Tap to focus before you shoot and skip photos with motion blur.
• The whole face in view. Front-facing or a slight turn, eyes open, and no sunglasses, masks or hair across the eyes.
• Even light. Soft window light or open shade outdoors beats harsh flash, deep shadows or strong colored light.
• A natural look. Heavy beauty filters, face-reshaping apps and photos already altered by another AI tool give the effect a distorted starting point.
• Framing that suits the effect. Head-and-shoulders for close-ups such as AI Crying, a selfie or half-body photo for Car Driving, most of the body for Ceiling Fan Transition, and a full-body or half-body photo for Storm.
• The original file. Use the photo itself rather than a screenshot or a copy compressed by a messaging app.

If you have no suitable photo, take a fresh one. Stand near a window or in open shade, ask a friend to shoot or use a timer, keep the camera at about eye level and take a few frames so you can keep the sharpest.

At the time of writing, the photo picker accepts JPEG, PNG, WebP or AVIF files under 10 MB. If the page shows different limits, follow the page.

Photos to avoid

• Group photos for single-person roles. Crop to one person, or better, use a separate solo photo.
• Collages, side-by-side images or photos with a second face in the background.
• Tiny faces in wide landscape shots.
• Very dark, backlit or heavily shadowed photos.
• Mirror selfies where the phone covers part of the face.
• Old, low-resolution photos that have been resaved many times.
• Anyone under 18, any public figure and anyone who has not agreed.

Matching photos for two-person scenes

In Raindance and Hotel Lobby both faces share one frame, so a much weaker photo stands out. Give both people similar framing, similar light and similar sharpness. Decide the roles before you upload: the first photo is always the person on the left. Name the files for yourself, for example "left" and "right", and check each thumbnail before you confirm. For three-person scenes, the Gas photo preparation guide explains role labels in more detail.

What a good photo can and cannot do

A clear photo reduces avoidable problems, but it does not lock in the outcome. Faces, hands and interactions can still vary, and exact likeness is not guaranteed. Each effect generates a new scene from your photo. It is not a face swap onto existing footage, and the result is not real footage of the people in it.

Photos you pick before signing in stay in that browser for up to 24 hours and upload only after you sign in, and you can delete them from the page at any time. Finished results stay in your history for 30 days.

The short checklist

• Every person is an adult (18+) who agreed to be included.
• You own the photo or have permission to use it.
• No celebrities, public figures or minors.
• One solo photo per person, in the right role.
• The face is sharp, fully visible and evenly lit.
• No heavy filters, sunglasses or masks.
• The framing suits the effect you chose.
• The file format and size match what the page accepts.

When your photos are ready, choose a scene in the AI video effects hub and check its status on the effect page before you sign in.