Can I use a photo of a friend or a celebrity?

A friend, only with their permission, and expect the same automatic checks your own photo goes through. A celebrity, or anyone who looks famous, is the photo most likely to be refused, and real or famous people are one of the six reasons Cinely's video provider can block a film. For a lookalike role, write an original character instead and let the AI draw the portrait.

Why does AI make my face smoother or younger than I am?

Some models lean toward polished, retouched-looking faces, and filtered selfies push them further. Remini reviewers describe an enhancer that turned faces plastic or doll-like and added makeup. Upload an unfiltered photo, write your age and details such as freckles, lines or facial hair into the character description, and avoid style words that suggest glamour.

Should I upload several photos instead of one?

Only if the tool asks for them and actually uses them. Some face apps want 10 to 20 selfies, and Lensa reviewers still describe results that varied from batch to batch. A mix of old and new photos, or very different lighting, can blur a likeness instead of sharpening it. One recent, sharp, evenly lit photo is a better start. On Cinely, you add one photo to your role.

Can I fix only the scene where my face looks wrong?

In many tools the only fix is another full-price generation. In Cinely's editor you can redo a single scene: "Edit current video" keeps the motion and changes only what you describe, while "Recreate from scratch" re-renders the scene from its original content with the same style and characters. Both cost a flat 60 credits, and your current movie stays unchanged until you save the new cut.

AI Video With My Face Looks Wrong? Here's Why

You uploaded a clear selfie and got back someone who looks like your cousin. Or you looked right in the first scene and like a stranger by the third. Most "AI video with my face" attempts go wrong for the same few reasons, and many of them are fixable before you spend another credit.

The short version: the AI doesn't paste your photo into the video. It redraws you in every frame and treats the photo as a strong hint. Anything that weakens the hint, like a new camera angle, harsh light, a crowded scene or a heavy art style, lets the model fill the gaps with a generic face. A better photo, fewer faces per shot and a style that suits your likeness solve a lot. Some photos, including your own, get declined outright.

Face do-overs also burn credits fast, as our guide to how AI video credits get spent shows.

What goes wrong when people put their face in AI video?

In September 2026 we read more than 12,000 one- and two-star reviews of 63 apps on the App Store, Google Play, Trustpilot and Capterra, about 1,500 of them for face and selfie apps. Behind a big share of those is one blunt question: why doesn't AI look like me? The complaints come in four kinds.

A stranger's face. About 17% of Lensa's low-star reviews say the results didn't look like the person. On Remini, face problems make up a fifth of this year's low-star App Store reviews, mostly about an enhancer that rewrites faces. Hailuo and VEED reviewers describe videos that looked nothing like the photos they uploaded.

A face that shifts. A Vidu reviewer on Google Play wrote that "if you upload a photo of yourself it will change the appearance regardless." A Lensa reviewer said the first set of avatars looked just like them, then added: "I used the exact same photos for the cosmic art option and they were awful."

The wrong skin tone or gender. A Kaiber reviewer on the App Store said the app "turned me, a light brown skinned, black-haired woman into a pale, red headed woman." On HubX's AI Video app, a parent wrote: "I Use my sons picture and it put him on female body !"

A photo that won't go through. Face and video apps flag ordinary selfies as inappropriate, ask for retakes again and again, or fail to detect a face at all.

Among video generators these are a small share of low-star reviews, but they're the main complaint of people telling a story across scenes. Models change often, so results vary.

Why does AI video with my face look like someone else?

Because a video model doesn't copy your face. It generates every frame from scratch, and your photo is one instruction among several.

When you add a reference photo, many systems reduce it to a compact description of identity: face shape, the spacing of your features, skin tone, hair. That description then competes with everything else in the request, including your prompt, the style, the lighting, the camera move and any other characters. When they pull in different directions, the model settles the conflict by drifting toward the kind of face it has seen countless times.

That's why fine details disappear first: eye shape, a crooked smile or a scar are small signals, and a generic face is a strong default. Tools also weigh the reference differently, which is why one photo can work in one app and fail in another.

Why does my face change between scenes?

Every clip is a new generation. In a multi-scene video the model redraws you from the reference each time, and small differences add up; a Vidu customer on Trustpilot described variations even with the same reference image. When the complaint is "the AI video changed my face," look for one of these six causes:

New angles. A front-facing selfie says nothing about your profile, so when the camera turns, the model invents what it never saw.
Lighting. Hard shadows, colored stage light and night scenes change how your features read.
Distance. In a wide shot your face is a small patch of pixels, too little to hold a likeness. Close and medium shots keep it.
Crowded scenes. Several reference faces compete, so features blend, faces swap, and people drop out; HubX reviewers describe people vanishing from group photos.
Heavy styles. Anime, painterly and 3D-cartoon looks each have a house face that pulls your features toward it, as the Lensa reviewer above found.
Big motion. Fast turns, laughter, or hands and hair across the face hide the features the model is matching.

Controlling those six is most of the work behind keeping AI characters consistent across scenes.

Why does AI change skin tone or gender?

Image and video models learn from huge photo collections that don't show everyone equally. When your reference signal is weak, the model fills the gap with the most common look for the words and style in your request. A role word like "princess," "Viking" or "CEO" carries its own default face, and lighting or color grading can shift skin tone on top of that.

About 5% of Lensa's low-star reviews describe changed skin tone or race, and 8 to 9% of FacePlay's describe whitewashed faces or swapped gender. A PixVerse reviewer on Google Play said the app kept changing characters to look Asian.

What helps:

Write down the traits it gets wrong. Skin tone, hair color and texture, age and gender, in the same words in every scene.
Upload a true-color photo. Filters and warm-tinted light change the skin tone the model reads.
Test in neutral light first. Check one daylight scene before night scenes or heavy color grading.
Swap loaded role words for details. If "princess" keeps changing your face, describe the dress and the palace instead.

Why was my photo rejected?

Because an automatic check can't tell who uploaded the photo. The feature that lets you star in your own movie would also let someone put a coworker, an ex or a celebrity into a video without asking, and a single image can't prove consent. So providers lean toward declining realistic, identifiable faces, especially ones that resemble public figures.

Rejections mean different things. A vague "inappropriate content" or policy message is the filter's judgment call. A request to retake the photo, or no face found, usually means the photo is too dark, too small or partly covered, and that you can fix.

Don't upload ten near-identical selfies hoping one slips through; they'll be judged the same way. Change the photo meaningfully once, and if that's declined too, use an illustrated or AI-drawn portrait. For what filters look for more broadly, see why AI flags harmless requests as content violations.

What kind of photo works best?

One good photo beats a dozen mixed ones. Check yours against this list before you upload:

Just you. No other faces, not even in the background, and no crop from a group shot.
A big, sharp, recent face. Head and shoulders filling much of the frame, in focus, with your current hair and glasses.
Soft, even light. Daylight from a window in front of you beats overhead bulbs, flash or colored light.
Straight on or slightly turned. Both eyes visible, with no sunglasses, low hat brim, hands or hair across your face.
The unfiltered original. Beauty mode, smoothing, screenshots and images squeezed by messaging apps all change or lose detail.
Plain clothes and background. No logos, text or costumes of famous characters.
Full body if you have one. A head-to-toe photo shows your build and outfit, so the model invents less.

Then keep scenes kind to the likeness: close shots for key moments, two or three people per shot, one style for the whole film. Our walkthrough on turning a photo into an AI movie character covers the setup.

How many people can keep their faces in one scene?

Fewer than most people expect. Before you plan a party scene, ask any "AI video generator with my face" tool how many faces it can lock in a single shot.

On Cinely, Standard (and Premium's default Pro mode) keeps up to 3 fixed faces in a scene, and Cinematic keeps up to 7. Characters beyond that appear as extras without a fixed face. A whole film can still hold up to 20 characters on the free plan and 50 on Premium, because faces are budgeted per scene, not per film.

Small casts get the strongest lock. Cinematic with 1 to 3 characters, or Standard and Pro with a single character, use "character assets" that also carry each character's own voice; larger casts fall back to photo references. A voice on the wrong person is a separate fix, covered in getting voices, sound and subtitles right. On the web, Cinematic is part of Cinely Premium.

How do you put yourself in a movie with AI on Cinely?

Cinely is an AI movie maker on the web, iOS and Android that lets you star in your own movie: your photo becomes the face reference for the scenes you appear in. If you searched "put yourself in a movie AI," here's the path:

Start on the create page. Preview is on by default there, so nothing renders until you approve.
Write your idea or paste a script, and add your photo to your role. You can be cast as the lead. If the photo isn't full-body, Cinely can redraw it into a head-to-toe image with the same face.
Review the free preview. Cinely writes the scenes and styles the characters. Roles without a photo get AI-drawn faces from their descriptions, with a counter such as "Faces ready 3/5".
Fix the cast before you pay. Edit any scene's prompt, choose which characters appear in each scene, regenerate, replace or re-describe any portrait, and pick voices. This is where you keep busy scenes under the face cap.
Handle a declined photo. If the provider's check rejects your photo, Cinely says it isn't a judgment about you and suggests another photo or an illustrated portrait. AI-drawn portraits can be declined too; then edit the description or add a photo.
Approve. If a role still has no face, a warning offers "Add photo", "Try again" or "Continue anyway", and a role you continue without appears with no fixed face. You pay for the scenes shown when you approve; discarding the preview costs nothing.

The honest limits: the provider declines many photos of real people, even your own, and face lock stops at 3 fixed faces per scene in Standard and Pro, or 7 in Cinematic. A reference guides your look but doesn't guarantee a perfect match in every frame. A scene that rendered with your face looking off isn't refunded, and redoing it in the editor costs 60 credits.

Start with one good photo and a two-scene test. Cast yourself in a short film on Cinely, check your portrait in the preview and your face in the first scenes, then go longer.