Best Photo for an AI Trend Video: What Works and What Fails
Oct 28, 2026 · AI Rap Generator
The single biggest factor in whether an AI trend video looks like you is the photo you upload. The video model builds your face from that one image. Give it a sharp, well-lit, front-facing face and it has plenty to work with; give it a tiny face at the edge of a group shot and it has to guess. This page covers how to pick the best photo for an AI trend video, before you spend any credits.
The six rules
1. Face the camera
Pick a photo where you look roughly straight at the lens. A slight angle is fine. A full profile is not: the model only sees half your face and has to invent the other half, so the result drifts towards a generic face.
2. Make the face big
Your face should fill at least a quarter of the photo. A head-and-shoulders selfie is ideal. A full-body photo from across the room usually puts the face at a few dozen pixels, which is not enough detail to carry through a 15-second video.
If your only good photo is a wide one, crop it on your phone first so the face is large, then upload the crop.
3. Keep it sharp
Motion blur, heavy compression from being forwarded through chat apps several times, and night shots with grain all remove the detail that makes a face recognisable. Use the original from your camera roll, not a screenshot of a story.
4. Nothing covering the face
No sunglasses, masks, hands over the mouth, or hair across the eyes. For lip-sync trends such as Hotel Lobby AI and Raindance AI the mouth matters most: keep it fully visible, ideally closed or relaxed.
5. One person per photo
Each slot is one person. If a photo has two faces, the model may pick the wrong one or blend them. For a group trend, upload a separate photo for each friend in their own slot.
6. Even, natural light
Daylight from a window, or a bright room, works best. Strong coloured lights, a hard shadow across half the face, or a beauty filter that smooths away features all make the result look less like you.
Common mistakes and quick fixes
| Mistake | What you get | Fix |
|---|---|---|
| Small face in a wide shot | Face looks generic | Crop so the face fills a quarter of the frame |
| Sunglasses or hat | Eyes and forehead invented | Use a photo without them |
| Two people in one photo | Wrong person, or a blend | One person per photo, one photo per slot |
| Beauty filter | Smooth, doll-like face | Use the unfiltered original |
| Profile or looking down | Face drifts | Pick a front-facing shot |
| Dark or grainy photo | Blurry face | Daylight photo from the camera roll |
Photos for group trends
Group templates such as Can Can Dance AI and Toc Toc AI take you plus up to two friends. The same rules apply to every photo, with one extra: photos taken in similar light look more natural together. Three selfies all taken outdoors in daylight blend better than one daylight selfie next to two night-out photos. The friends guide covers group videos in more detail.
Pets and other faces
Some meme templates, such as ABCDEFG I Need Money AI, work with a pet as the face. Use a front-facing shot where the animal looks at the camera and the mouth area is visible.
Photos you must not use
Use only photos of yourself or of adults who agreed to be in the video. Never upload photos of children, celebrities, politicians or strangers. These are blocked by our acceptable use rules, and you confirm this with a checkbox before every video.
What happens to the photo after upload
Your photo is stored privately, sent to the video model only for your request, and deleted automatically 24 hours after upload. It is never used to train models.
Check before you press Generate
The preview in each slot shows exactly what the model gets. Before you press Generate, look at it once more: big face, sharp, front-facing, nothing covering it, one person. If the preview passes, the photo is ready. Head to all templates to pick your trend.