How to make AI photos of yourself
To make AI photos of yourself, start with one clear, well-lit photo of your face, add it to an AI image model that can edit a photo while keeping the person recognisable, then pick a ready-made style or describe the scene you want. Current models such as Nano Banana Pro, GPT Image 2.5 and Seedream 4.5 work from a single reference photo, so there is no training step and a result usually arrives within a minute or two. The quality of that one photo decides, more than anything else, how much the result looks like you.
Start with the right photo
Every AI portrait is built from the reference photo, so a few seconds spent choosing it pays off more than any prompt. The model can only keep what it can see. If half your face is in shadow or hidden behind sunglasses, it has to guess, and guesses are where the likeness slips.
- Your face is fully visible and roughly facing the camera, with no sunglasses, mask or hand in the way.
- Soft, even light, such as daylight from a window. Avoid hard shadows and coloured party lighting.
- Only you in the frame, or a crop where you are clearly the main subject.
- A sharp, recent photo straight from the camera, not a screenshot or a photo with a beauty filter.
- A neutral or natural expression, unless you want that exact expression kept.
- For full-body scenes, a second photo that shows your whole outfit and build.
Step by step
The workflow is the same in most apps built on modern image models:
- Choose a starting point: a ready-made style (a headshot, a travel scene, a retro look) or a blank prompt.
- Add your photo. If the app accepts several reference images, add a second angle of your face, or a photo of an outfit or place you want in the picture.
- Pick the model. Use a detail-focused model for close portraits and a fast one for trying ideas.
- Set the aspect ratio for where the picture will go: 4:5 or 3:4 for an Instagram post, 9:16 for a Story, 1:1 for a profile picture.
- Generate, then refine. Change one thing at a time ('make the light warmer', 'zoom out to show the whole jacket') instead of rewriting everything.
- Save the version you like at the highest resolution offered.
Use a style or write your own prompt
Styles are the quickest route. Someone has already written and tested the prompt and chosen the model, so all you add is the photo. They are ideal for well-known looks such as professional headshots, holiday cards or a retro decade, and for joining a trend without working out how it was made.
Writing your own prompt gives you control over everything else: the place, the outfit, the light, the camera. Describe the picture in full sentences and state what must not change. A prompt like 'Keep my face, hair and skin tone exactly as in the photo. Show me on a balcony in Lisbon at golden hour, wearing a linen shirt, photographed with an 85mm lens and a softly blurred background' gives the model a clear brief. The AI portrait prompts page has examples to copy and adapt.
In Luren AI you can do both. The styles, trends and collections fill in the prompt for you, while the Image tab lets you type your own prompt, add up to six reference photos and choose the model yourself.
Which AI model to choose
Models are good at different things, which is why apps that offer several of them let you switch per image. As a rule of thumb:
- Nano Banana Pro (Google): the most detail and the most consistent faces, at 2K or 4K. Good for close portraits and images with readable text.
- Nano Banana 2 (Google): fast and strong at keeping your face while it changes the scene, outfit or light, up to 4K.
- GPT Image 2.5 (OpenAI): precise edits that change only what you ask for, useful when you refine one image over several rounds.
- Seedream 4.5 (ByteDance): natural, photographic results. It reads many reference photos at once, which helps when combining you, an outfit and a place.
- Grok Imagine (xAI): bold, stylised looks at speed.
How to make it look like you
Most likeness problems come from a prompt that asks for more change than the photo can support. If you describe a different age, hair colour or face shape, the model will follow you and the person will drift. Describe the scene and the clothes, and leave your features to the photo. You can compare what each model does best on the AI models page.
- Say explicitly that the face must stay the same, and mention hair and skin tone if they matter to you.
- Keep the framing close to your photo. A head-and-shoulders selfie makes a better close portrait than a full-body action shot.
- Add a second photo of your face from another angle if the model accepts several references.
- Generate two or three versions. The likeness varies between runs, and one is usually clearly the best.
- For couple or group pictures, add one clear photo per person and say who stands where.
Privacy and fair use
Only use photos of yourself, or of people who have agreed to it. A photo taken on a phone can also carry the exact GPS location where it was taken, so check it before you upload or share it. The free remove photo metadata tool shows what a photo reveals and saves a clean copy without uploading it anywhere.
Treat AI portraits as what they are. Images made with Google's Nano Banana models carry SynthID, an invisible watermark that identifies them as AI-generated, and an AI picture is never a substitute where a genuine photograph is required, such as an ID card or a passport application.
