How to Make AI Images of Yourself From One Photo

To make AI images of yourself, upload one or more clear photos of your face to an image-editing model and describe the new scene, outfit or style you want; the model keeps your features and repaints everything else. You no longer need to train a personal model on 20 selfies. On VdoBloom this takes a single photo in the Image Edit tab and costs 3 to 24 credits per image, depending on the model you pick.

Disclosure: VdoBloom is our platform.

How "AI images of me" work now (and why you do not need to train anything)

Two or three years ago, getting yourself into an AI image meant fine-tuning a model on a folder of your selfies, waiting, and hoping it learned your face. Most current tools skip that. They use reference-image editing: you give the model your photo as input, write what should change, and the model generates a new image that is anchored to the face, hair and build in that photo.

This matters for three common questions:

• How many photos do I need? One sharp, front-facing photo is enough for most edits. Some models accept several reference images, which helps when you want a different angle or full-body shot.
• Is it the same as a face swap? Not quite. A face swap pastes your face onto someone else's body in an existing photo. Reference editing builds a new picture of you. Face swap has its own rules and a separate guide: how to face swap a photo well.
• Will it look exactly like me? Close, not perfect. Expect small drift in eye shape, jawline or skin texture, especially in wide shots and heavy stylization. The fixes are covered below.

Step by step: put yourself in any scene

1. Pick the right source photo. Face clearly visible, even light, no sunglasses, no heavy filter, nobody else in the frame. A phone photo taken by a window works better than a dim party shot.
2. Open the editor. On VdoBloom go to Image Edit, tick the consent box (it confirms the photo is of an adult and that you have the right to use it) and upload your photo. Files up to 10MB are accepted.
3. Write what changes, and say what must stay. Example: "Keep the person's face, hair and skin tone identical to the reference. Place them on a rooftop in Tokyo at night, wearing a black trench coat, neon reflections, 35mm photo, shallow depth of field."
4. Choose a model and resolution. The default Auto option works for most scenes. Details on models below.
5. Generate, compare, iterate. Change one thing at a time (lighting, outfit, camera distance). If the face drifts, shorten the prompt and move the camera closer.

Prompt patterns that hold likeness better

• Say "the person in the reference image" instead of describing your features. Describing "brown eyes, short hair" invites the model to invent a new face that matches the words.
• Ask for a medium or close shot first. Full-body shots in big scenes shrink the face to a few pixels, which is where likeness is lost.
• Keep style words modest for realistic results. "Cinematic lighting" is fine; stacking five art styles pushes the model away from your face.
• If you want a different outfit, describe the outfit, not the body.

Which VdoBloom tab to use for which result

Several tabs are really the same idea (your photo plus a tuned prompt) packaged for a specific job. Picking the right one saves prompt writing.

The headshot and try-on tabs both run on Nano Banana Edit at 5 credits per image. The Headshots tab makes 4 images per style pack, so the three packs selected by default (Corporate, Casual Lifestyle, Studio) come to 12 images and 60 credits. Images that fail are refunded.

Choosing a model in Image Edit (and what it costs)

Credit prices below are the current per-image prices on VdoBloom, as of September 2026.

• Auto (default) - 7 credits. Routes the edit for you and has fewer restrictions on things like swimwear or fashion shots.
• Seedream 4.5 Edit - 7 credits. Accepts up to 10 reference images; good for combining your photo with a location or product. See the model page.
• Nano Banana Edit - 5 credits. Strong at keeping a face stable across edits, but it applies strict filtering.
• Nano Banana Pro - 18 credits at 1K or 2K, 24 credits at 4K. Use it when you need a large print or a crisp 4K portrait.
• GPT Image 2 - 3 credits at 1K, 5 at 2K, 8 at 4K. Cheapest way to try ideas; strict filtering.

Worked examples: one travel-style picture of yourself on Auto is 7 credits. Ten quick drafts on GPT Image 2 at 1K are 30 credits, and a final 4K version on Nano Banana Pro adds 24. New accounts start with 10 free credits, which covers one Auto edit or two Nano Banana Edit images, enough to see whether the likeness works for your face before you pay.

Making it look like you, not a cousin

Likeness drift is the number one complaint with any "AI photo of me" tool, including ChatGPT, Gemini and dedicated apps. It is a limit of the technology, not something one setting fixes. What helps:

• Better input beats longer prompts. Retake the source photo in daylight before rewriting the prompt for the fifth time.
• Use two references for tricky angles. A front photo plus a three-quarter photo gives the model more to work with for profile shots.
• Edit in two passes. First generate the scene with a close camera. Then ask for "the same image, zoom out to show full body" using the result as the new reference.
• Keep glasses, beard and hairstyle consistent with the source unless changing them is the point.
• Generate several and pick. The Headshots and Pose Variations tabs return multiple images per run for this reason.

If you plan to turn these stills into video with the same face, the same rules apply across clips. Our guide on keeping a character consistent in AI video covers that side.

Free options, ChatGPT, and phones

People often ask how to do this for free or inside apps they already use. As of September 2026:

• ChatGPT and Gemini both let you upload a photo and ask for an edited version in chat, with daily limits on free plans. Both apply strict rules on real people, and results can be refused when a prompt looks like it targets someone else.
• Social apps such as Instagram and Facebook add and remove AI selfie features by country and account, so check inside the app rather than trusting a months-old tutorial.
• Free tiers everywhere are rate limited. On VdoBloom the 10 signup credits are the free tier; there is no daily refill.

The trade-off with paid multi-model tools is choice: you can switch models when one keeps drifting on your face or refuses a harmless outfit.

Consent and safety: only your own face

These tools make it easy to generate convincing pictures of any person, so the rules are short and firm:

• Use your own photos, or photos of people who have agreed to it.
• Never upload photos of minors. VdoBloom requires you to confirm every uploaded person is 18+ and blocks content involving minors.
• No sexual or intimate images of real people. Non-consensual intimate images, including AI-generated ones, are illegal in many places, such as under the US TAKE IT DOWN Act.
• No public figures or other people in content designed to deceive, such as fake endorsements or fake news photos.

VdoBloom runs its own moderation on image uploads and generations, and blocks non-consensual and underage content.

Frequently asked questions

How do I make AI images of myself for free?

Use a free allowance: ChatGPT and Gemini have limited free image generation, and VdoBloom gives new accounts 10 credits, enough for one Auto edit (7 credits) or two Nano Banana Edit images (5 credits each).

How many photos do I need to create an AI image of myself?

One clear photo is enough for reference-based editors. Add a second photo from another angle if you want profile or full-body shots to hold your likeness.

How do I make AI images of yourself with ChatGPT?

Upload a photo in a chat and describe the scene, saying the face should stay the same. It works for many people, but likeness can drift and real-person rules are strict, so try a few versions.

Can I make an AI video of myself from these images?

Yes. Take your best still into an image-to-video tool and animate it. Starting from a good still is more reliable than asking a video model to invent your face.

Why does the AI image not look like me?

Usually a small or blurry face in the source, a wide shot, or a prompt that describes your features instead of referencing the photo. Use a sharper photo, a closer camera and "keep the person identical to the reference".