How to Make a Picture Full Body with AI (Step by Step, 2026)

To make a picture full body with AI, you give the photo a taller canvas and let an outpainting model paint the missing torso, legs, feet and floor below it. That is what an image extender does. On Kenerate you do it in Expand Image: grow the bottom edge (up to 100% per run, so a head-and-shoulders crop usually needs two or three passes), then upscale the result. If the crop is too tight to extend cleanly, the other route is to regenerate a new full-body shot with your photo as a reference.

Updated 20 October 2026. Credit prices and limits below come from Kenerate's own tool configuration as of that date.

One thing to know before you start: the AI does not "find" the rest of the body. The legs, shoes and pose below your crop line are invented to match what is visible. That is fine for an avatar, a character sheet, a mock-up or a social post. It is not a record of what someone looked like, and you should only do it to photos of yourself or people who have agreed to it.

What does "make this picture full body" actually mean?

People search this phrase when they have a photo that stops at the chest or waist and want one that shows a standing person head to toe. There are two different jobs hiding behind it:

1. Extend the same photo. Keep every original pixel and add new image below (and usually beside) it. This is outpainting, the technique OpenAI introduced to DALL·E in August 2022 as a way to continue an image "beyond its original borders" while respecting shadows, reflections and textures (OpenAI).
2. Make a new full-body picture of the same subject. Generate a fresh image from a prompt and use your photo as a reference so the face, hair and outfit carry over. Nothing from the original frame is kept pixel-for-pixel.

Most of the time you want the first one, because it keeps the face exactly as it was. The second is the fallback when the original crop gives the model too little to work with.

Outpainting vs regenerating: which should you use?

A simple rule: if the photo already shows the hips, extend it. If it stops at the shoulders and you need a confident standing pose, regenerate with a reference, then fix details with an edit.

How much do you need to extend a cropped photo?

This is where most first attempts go wrong. Figure-drawing convention puts an adult at roughly seven to eight head heights tall. A head-and-shoulders crop shows about two of those; a waist-up crop shows about three to three and a half. So to reach the feet you are asking the model to invent:

• Head-and-shoulders photo: roughly three times the photo's current height below it.
• Waist-up photo: roughly one and a half times its height.
• Mid-thigh photo: roughly half its height.

Kenerate's Expand Image lets each edge grow by 0 to 100% of the photo's own height or width per run. That means:

• A mid-thigh photo can usually reach the feet in one run (bottom +50% to +60%).
• A waist-up photo needs two runs (bottom +100%, then extend the result again).
• A head-and-shoulders photo needs two or three runs, and is often better regenerated instead.

Going in steps is not just a limit of the tool. Extending a busy image in one giant jump tends to look less convincing than several smaller ones, and generative models find hands and complex detail harder to continue than open space (MagicShot's outpainting guide makes the same point). Each pass gives the model a fresh, larger context to work from.

How do you make a picture full body on Kenerate, step by step?

Here is the outpainting route using Kenerate's AI image extender.

1. Open Expand Image and add the photo. Upload a JPG, PNG or WebP up to 25 MB, or paste a public image link. Running it needs an account with a verified email; new accounts get 25 free credits, which covers one run on any of the expand models.
2. Switch to Grow edges. The tool has two modes: Ratio (pick one of 10 target shapes) and Grow edges (add a percentage to each side). For a full body you want Grow edges, because ratio mode only adds the smallest frame that fits the ratio. Going from a 4:5 portrait to 9:16 adds only about 42% height, which is rarely enough to reach the feet.
3. Grow the bottom edge. Set bottom to 100% for a waist-up photo (or 50% to 60% for a mid-thigh one). Add 10% to 25% on the left and right as well: a standing figure needs some room for elbows and a wider stance, and a little extra floor reads more naturally than a body touching the frame.
4. Pick a model. Six models are offered. Nano Banana 2 Fast is the default and the cheapest at 13.5 credits per run. See the model table below.
5. Describe the new area (optional, up to 400 characters). Without a note, the model continues whatever naturally lies beyond the edge. With one, you steer the outfit and setting. Examples are in the prompt section below.
6. Run 1 or 2 variations. Two variations cost twice the credits but make it easier to pick a natural-looking pair of legs. The result shows an outline of where your original sits.
7. Repeat if the feet are not in frame yet. Download the best variation, or send it straight back into Expand Image, and grow the bottom edge again.
8. Upscale the final image. The working canvas is capped at 2,048 px on the long side and each model answers at its own resolution, so a multi-pass full-body image will be smaller than a modern phone photo. Run it through the AI image upscaler at the end, not between passes.

If you want to understand the technique itself in more depth, the outpainting explainer covers how the canvas and mask are built.

Which AI model is best for extending a photo to full body?

These are the six models available in Kenerate's Expand Image as of 20 October 2026, with the price of one run (one variation) from the tool's credit calculator.

A sensible way to spend credits: draft the first pass with Nano Banana 2 Fast, and if the clothing or proportions look off, re-run that same pass with Nano Banana 2.1 or GPT Image 2.5 Sunburst rather than pushing on to the next pass with a flawed base. Mistakes compound: a slightly short torso in pass one becomes very short legs in pass two.

On credits: Kenerate's smallest top-up is $15 for 1,300 credits, which works out to about 96 runs of Nano Banana 2 Fast. A three-pass extension with two variations each is 81 credits on that model.

What should you write in the prompt?

The note describes only the new area, not the whole picture. Keep it short and concrete. Templates that work:

• Casual outfit: "dark blue straight-leg jeans and white sneakers, standing on a light wooden floor, feet slightly apart"
• Formal: "charcoal suit trousers matching the jacket, black leather shoes, standing on grey stone pavement"
• Studio portrait: "plain light grey studio backdrop continuing to the floor, soft shadow under the feet"
• Outdoor: "continue the park path and grass, legs in khaki chinos, brown boots, relaxed standing pose"
• Dress or skirt: "knee-length black dress continuing from the top, bare legs, black flats, city sidewalk"

Tips that make the biggest difference:

• Name the clothing for the lower half. If the photo shows a jacket, the model has to guess the trousers. Telling it removes the most common mismatch.
• Name the floor or ground. Feet look far more natural when they stand on something specific.
• Describe the stance, not anatomy. "Standing, weight on one leg" works better than listing body parts.
• Keep it consistent across passes. Use the same note on pass two so the outfit doesn't change halfway down.

Why do AI full-body extensions look wrong, and how do you fix them?

Hands are a special case. If your original crop slices through the hands, the model has to finish fingers it can only partly see, which is where AI struggles most. Our guide on how to fix AI hands and fingers covers the repair once the body is extended.

When should you regenerate instead of extending?

Extending stops being the best option when:

• The crop is a tight headshot. Three passes of invented body are more likely to drift in proportion than one clean generation.
• You want a different pose. Outpainting continues the visible pose. If the arms are crossed in the photo, they stay crossed.
• The original is low resolution. Every pass inherits the blur; a regenerated image starts fresh.

To regenerate on Kenerate, open Image Studio, add your photo as a reference, and choose a reference-capable model such as Nano Banana 2.1, which accepts up to 14 reference images and costs 21 credits at 1K (31.5 at 2K, 42 at 4K). A prompt that works: "Full-body photo of the same person from the reference image, standing, head to toe in frame, wearing [outfit], [setting], natural light, same face and hairstyle." Generate a few, keep the one whose face matches best, then use Expand Image if you need more space around it.

Other vendors land in the same place. OpenArt's full-body generator, for example, is an image expander with adjustable canvas and an optional prompt, and recommends starting from a clear, high-resolution input (OpenArt).

Which aspect ratio should a full-body picture be?

A standing person is tall and narrow, so full-body images are almost always portrait:

• 2:3 suits a single standing figure with a little floor and headroom. It is also the poster and Pinterest shape.
• 9:16 is for Stories, Reels and TikTok covers. It leaves extra space above or below, which you may want to fill with background.
• 4:5 is Instagram's feed portrait. It is a bit squat for a full figure; you often end up with the feet close to the bottom edge.
• 3:4 is a standard phone portrait and works for a full figure if you leave the sides narrow.

Once the body is complete, you can switch to Ratio mode in Expand Image to land on an exact platform shape, with the photo anchored to the top or centre. If you need a different direction entirely, such as turning a portrait into a wide banner, the expand image AI page covers that use case.

Is it okay to make a full-body version of someone's photo?

Use it on your own photos, on characters you have made, or on people who have clearly agreed to it. A full-body extension invents a body that the camera never captured, and presenting it as real can mislead people. Kenerate moderates the note you write for the new area, but the responsibility for whose photo you upload is yours. If you are making an avatar or a character design, a consistent made-up character is usually the better starting point anyway.

Sources

• OpenAI, "DALL·E: Introducing outpainting" (31 August 2022): https://openai.com/index/dall-e-introducing-outpainting/
• MagicShot, "AI expand image: extend photos beyond the frame" (outpainting guide): https://magicshot.ai/blog/ai-expand-image-extend-photos-beyond-the-frame-outpainting-guide
• OpenArt, "AI Full Body Generator": https://openart.ai/features/ai-full-body-generator
• Kenerate product facts: Expand Image limits and credits (src/lib/expand-image.ts), model specs (src/lib/models-config.ts), plans (src/lib/pricing.ts), as of 20 October 2026.