# AI image part extractor

Take one object out of a picture and make it on its own. The AI image part extractor lists the parts it finds, or you name one — a weapon, a helmet, a vehicle, anything you type — then redraws it whole on white, filling in what was hidden, or cuts it out of your picture's own pixels. Each part can become its own 3D model, all of them in one batch.

> Canonical: https://ai3d.studio/image-part-extractor

## Three parts redrawn from one knight

The knight that Try with a sample character loads, with one redraw per part in Nano Banana 2: 10 credits a part, nothing retouched. Each part is completed where the picture hid it and set alone on white for image to 3D.

### Helmet

The winged helmet as empty headwear, open where the face was, with the plume and the gold trim kept.

### Cape

The cape on its own, completed where the knight’s body hid it, with both gold clasps at the shoulders.

### Boots

Both boots side by side, completed where the cape and the stance hid them.

## How the image part extractor takes a part out

Each part is redrawn whole or cut out of your picture, as you choose.

### Parts found for you

Analyze this picture costs 1 credit and lists up to 12 parts, each with a name, a line of description and a box drawn on your picture.
Rename a part, untick it or add one by name, then choose Redraw or Cut out for each one before anything is extracted. Names are capped at 40 characters.

### Hidden areas completed

Redraw rebuilds the whole part for 10 credits. Worn on the knight, the helmet is half face; redrawn by Nano Banana 2 it comes back as headwear on its own, the visor open and the inside finished where the head used to be.
The redraw prompt keeps the part's shape, proportions, colors, materials and style.

### Cut out, exact pixels

Cut out finds the named part with SAM 3 and keeps only its own pixels on a transparent PNG, cropped to the part, for 1 credit. Nothing is redrawn, so nothing is invented.
On the knight that means an empty gap where the face was, and a plume trimmed where it touched the background: exact pixels, and only those.

## The image part extractor in numbers

### Up to 12 parts found for 1 credit

Analyze the picture once and get a list of up to 12 parts, each with a name, a one-line description, a box on your picture and a note when something hides it.

### 11 part presets, or your own name

Head, outfit, boots, weapon, shield, helmet, cape and bag from a character; the main character, a vehicle or a building from a scene. Or name any part in up to 40 characters, from a lantern to a dragon's wing.

### Redraw 10 credits, cut out 1 credit

A redraw is one Nano Banana 2, GPT Images 2 or Seedream 5 Lite edit at the default size; Nano Banana Pro is 12. A cut-out is one SAM 3 segmentation of your own picture. Choose per part.

### Several parts per go, then one batch to 3D

Pick several parts and each comes back as its own picture, one run per part, started as your plan allows. Make 3D for the parts then sends them to Image to 3D as one batch, on a plan with batch.

## What the image part extractor can take out

Presets for the common cases, and your own words for the rest.

### Gear from a character

Weapon, shield, helmet or hat, cape or cloak, bag or backpack.

### Subjects from a scene

The main character without their props, a vehicle, or the main building.

### Anything you can name

A lantern, a left boot, a dragon's wing, a chair: one short name per run.

### Ready for 3D

A single object, whole and centered, on white from a redraw or transparent from a cut-out, is what Image to 3D and 3D printing prep need.

## Split a picture in three steps

Upload, find the parts, extract them.

### Upload the picture

A character with props or a scene with objects, up to 10 MB.

### Find or name the parts

Analyze the picture and review the list it returns, or pick presets such as Head and Weapon, or type your own, like lantern.

### Extract it

Choose Redraw or Cut out for each part and run it, then send one part to Image to 3D or make 3D for every part in one batch.

## Limits of the image part extractor

A redraw imagines what is hidden; a cut-out keeps only what is visible.

### Hidden areas are imagined

With Redraw, what was behind other things is drawn to match, not recovered; check the result before printing it. A cut-out does not fill it in at all.

### Each part is a paid run

Three parts cost three runs, 10 credits for a redraw and 1 for a cut-out; the Extract button shows the total before anything starts.

### Name it clearly

If two things fit the name, the model picks one; name the part by color or side, like left boot.

### A cut-out can find nothing

If the segmentation finds nothing that matches the name, or finds the whole picture, the run fails and its credit is returned; try that part with Redraw. Cut-outs work at up to 2,048 px on the longer side.

### The part list is a first pass

The analysis can miss a part or join two, and it lists at most 12. Rename, untick or add parts by name; a redraw reads the names, not the boxes.

### Batch comes with some plans

Make 3D for the parts is the Image to 3D batch, which only plans that include batch can start; the form names them before anything is charged.

## After the image part extractor

From one picture to a set of 3D parts.

### [Make each part 3D](https://ai3d.studio/image-to-3d)

Image to 3D builds a textured model from each extracted part, for credits.

### [Split a 3D model instead](https://ai3d.studio/3d-model-segmentation)

3D model segmentation divides a finished model into parts.

### [Pose the character](https://ai3d.studio/ai-pose-changer)

The AI pose changer puts the main character in a T-pose for rigging.

## Questions about the image part extractor

Answers about parts, hidden areas and cost.

### What does the AI image part extractor do?

The AI image part extractor takes one part out of a picture — a weapon, a helmet, a vehicle or anything you name — and either redraws it whole and alone on a white background with one image edit or cuts it out of your picture's own pixels, ready to become its own 3D model.

### Can the image part extractor find the parts for me?

Yes. The image part extractor can analyze your picture for 1 credit and list up to 12 separable parts, each with a name, a one-line description, a box drawn on your picture and a note when something hides it. You rename, untick or add parts, and only the parts you keep are extracted.

### Can the image part extractor recover a part that is partly hidden?

Yes, with Redraw. The image part extractor redraws the part rather than cutting it out, so areas hidden behind a hand or another object are filled in to match; they are drawn to fit, not recovered from the picture. A cut-out keeps only the visible pixels.

### How do Redraw and Cut out work in the image part extractor?

Redraw asks an image model to draw the part again, whole and alone on white, so it can fill in what is hidden. Cut out asks SAM 3 to find the part in your picture and keeps only its own pixels on a transparent background, so nothing is invented and nothing hidden returns. The image part extractor lets you choose per part and mix both in one run.

### How do I get several parts with the image part extractor?

Pick every part you want — say Head, Outfit and Boots — or keep the ones the analysis listed, and the image part extractor runs one edit per part from the same picture. Each part is its own result with its own Turn into 3D action.

### How much does the image part extractor cost?

Each part in the image part extractor is one run: a redraw is 10 credits on Nano Banana 2, GPT Images 2 or Seedream 5 Lite at the default size and 12 on Nano Banana Pro, and a cut-out is 1 credit. The optional picture analysis that lists the parts is 1 credit.

### Can the image part extractor make 3D models of all the parts at once?

Yes, on a plan with batch. After the parts are extracted, Make 3D for the parts sends every part picture that came out to Image to 3D as one batch, one 3D model each, started on our servers so it keeps running with the tab closed. The image part extractor shows parts times the model's credits and names the plans first; a part that did not extract is left out.

### Who uses the AI image part extractor?

The AI image part extractor suits game artists building a kit of props from concept art, and makers who want to print a character's helmet or sword as a separate piece.

### How many credits does Image Part Extractor use?

Image Part Extractor draws 10 credits per run.

### What limits should I check before using Image Part Extractor?

Generated output from Image Part Extractor can contain visual, audio, or factual errors. Review each result before publishing, and upload only material you are authorized to use. Where the form accepts a reference image, supported files are JPEG, PNG, WebP up to 5 MB each.

### What happens to the files I upload to Image Part Extractor?

An upload is stored against your own account, not pooled with other people's. Deleting the account removes it: the control is in Settings under Security, the request is confirmed by email, and the erasure removes the stored objects it owns. How long anything else is kept, and which processors handle it, is stated in the privacy policy.

### Can I use what Image Part Extractor makes commercially?

Every plan, the free tier included, lets you use what Image Part Extractor makes commercially: the outputs generated for you are yours.

### How does Image Part Extractor compare, and what alternatives fit a different workflow?

Image Part Extractor: AI image part extractor: isolate a weapon, helmet, cape or any part you name from a picture, completed on white, then turn each part into its own 3D model. Related options: AI 3D model generator with 11 named models: start from one image, a prompt, up to eight views or a sketch. From 13 credits; GLB, FBX, OBJ, STL or USDZ out. Sketch to 3D with Hunyuan3D Sketch: upload a line drawing, add a short description, and get a textured GLB or OBJ model for 47 credits. Image to prompt: upload a picture and get up to 4 prompts for an AI image generator or Text to 3D, in 6 styles and 11 languages, for 2 credits. AI 3D model comparison: Compare AI 3D models side by side: Tripo, Hunyuan3D, TRELLIS.2, Rodin, Meshy, Hitem3D and Seed3D by credits per image run (29 to 263), inputs, formats and time. Sketch to image AI: draw in your browser or upload a sketch, render it as a 3D render, game asset, clay figure or photo, then turn it into a 3D model. Input image guide: Prepare an image for image to 3D: one subject, fully in frame, a plain background, soft light, and a JPEG or PNG under the model's size cap. Four checks. Choose by the input each one needs, the output it delivers, its controls and limits, whether it is available now and its credit cost; each is linked under Related tools on this page.

## Take a part out of your picture now

Upload a picture, name the part, and make it its own 3D model.

## Related tools

### [AI 3D model generator](https://ai3d.studio/ai-3d-model-generator)

### [Sketch to 3D](https://ai3d.studio/sketch-to-3d)

### [Image to Prompt](https://ai3d.studio/image-to-prompt)

### [AI 3D model comparison](https://ai3d.studio/compare-ai-3d-models)

### [Sketch to Image](https://ai3d.studio/sketch-to-image)

### [Input image guide](https://ai3d.studio/prepare-image-for-3d)

_Last updated October 19, 2026_