Plus and Pro plans only. AI Image to 3D in Event Design App turns a single photo of one object, such as a chair, sofa or prop, into a 3D model (turn a photo into a model) you can rotate, resize to its real-world height and download as a GLB file. It spends credits on each generation. A plain background and a whole object give the best result.
The catalog doesn't have the exact lounge sofa your client is renting. AI Image to 3D builds a model of it from a single photo, so you can set it to its real size and use it in your plans.
This article covers choosing the photo, following the job, and sizing and downloading the result. To keep the model for reuse, see Save generated models to the catalog.
📺 Choose a Source Image and Generate an Image-to-3D Model, Use Image-to-3D History and Recover from Job States and Set Model Height and Download a GLB — three short lessons, about 4 minutes each.
AI Image to 3D needs a Plus or Pro plan (the Free plan can't use it) and spends credits for each generation. Three things have to be true, and they are separate:
Access. Your plan (and company settings, if you're in a company) include it.
Credits. You have enough for the request.
A valid source image. Generate 3D Model stays disabled until you add one. A disabled button on an empty draft doesn't mean you're out of credits.
⚠ The screen doesn't show a price. There is no per-request cost or confirmation step on the upload screen. Check your balance in the account menu and confirm the current cost before you click. Costs can change.
The photo decides most of the result. Use one object on a simple or solid-color background, with a clear outline and no text overlays. The object should be larger than half the frame, with the whole thing visible.
Good source | Poor source |
|---|---|
One sofa on white, fully in frame | A staged scene with tables, chairs, flowers and a canopy |
Sharp edges, no logos or captions | Text overlays, watermarks, blended backgrounds |
Subject fills most of the frame | A small object in a big room |
File requirements: JPG, PNG or WebP, up to 10 MB. A wrong format or an oversized file is rejected on upload (Only JPG, PNG, and WebP images are allowed / Image size must be less than 10MB). The screen lists the current requirements, so recheck them when you prepare a new file.
[SCREENSHOT 1 — good vs poor source image side by side]
To start a generation:
Open the Catalog panel and choose AI Image to 3D.
Click Upload Image and select your file. The preview and file name appear, and the draft is marked Unsaved. That's normal: the image is staged for generation.
Check that the preview shows the right object with good framing.
Click Generate 3D Model once.
The button shows Starting worker… while the image uploads and the job is created. Credits are charged when the job starts.
⚠ Click once, then follow the existing entry. A second click starts a second job and a second charge.
[SCREENSHOT 2 — upload screen with preview and Generate 3D Model]
The History list on the AI Image to 3D screen shows every generation. Scroll to History. Each request is identified by its source thumbnail, file name and submission time. Match those, especially when you have several similar generations.
State you'll see | What it means |
|---|---|
Queued | Waiting to start. |
Processing with Preview Unavailable | Work is underway. There is nothing to open yet. |
Completed with Preview Ready | The generation finished. Open it. |
Failed | It didn't finish. Open the entry and read the reason. |
Saved | You've already saved it to the catalog. |
⚠ Preview Unavailable is not Failed. While a job is processing, no preview exists. Wait for the state to change. Don't submit a duplicate.
You don't have to stay on the screen. If a generation finishes while you're elsewhere in the editor, a green AI Image to 3D ready notice appears with a prompt to open it.
The empty Current Draft on the left is for a new upload. It doesn't remove anything from History. Click a History entry to inspect it.
[SCREENSHOT 3 — History with a Processing entry and a Completed entry]
Open the completed History entry. When History Preview loads, rotate the model. Check the seat back, arms and base from another angle. Completed means the generation ran, not that the geometry is right. Compare the shape with your photo. If the model is wrong, a new source image is the fix, not scaling.
A generated model arrives at an arbitrary size. In Model Transform (the panel that resizes the generated model):
Check the units toggle: Imperial (inches) or Metric (centimeters).
Note the detected height, about 19.85 in the sofa example. Don't treat it as the real size.
Type the real height in Target Height and leave the field. The sofa example uses 43 inches.
The scale factor updates (about 2.166 for the sofa). There is no Apply button.
The same factor applies in every direction, so proportions stay consistent. The preview may stay about the same size on screen, so judge the size by the target height and scale factor, not by how big it looks.
⚠ Inches and centimeters are different heights. 43 in is 109.22 cm. If a model comes out huge or tiny, check the unit label and decimal point first. Correct the target before downloading.
⚠ Scaling doesn't repair the shape. It resizes the model. It won't fix a distorted surface or add a detail the generation missed.
[SCREENSHOT 4 — Model Transform with Target Height 43 and scale factor]
A GLB is a standard 3D model file format. Click Download 3D Model. The button shows Preparing GLB… and then downloads a .glb file at your target height. In the lesson, the downloaded sofa measured 1.0922 m, which is 43 inches.
Name the file so you can find it later and note the height, for example Sofa 43 inch v1.
When an AI Image to 3D job fails, read the entry's final state and message first. A failed job shows Failed and a message such as Generation failed. Try a different image or try again later. There isn't a Cancel button on this screen.
Credits are charged at the start. A refund is issued when a job ends as failed or cancelled, shown as a Feature Refund in your credit history. A job that completes is not refunded, even if you don't like the model, because the generation ran. Confirm any unclear adjustment in your credit history or with your account administrator. See Troubleshoot AI jobs.
Pick a photo of one object on a plain background. Check format and size.
Upload it, check the preview, and click Generate 3D Model once.
Find the entry in History by thumbnail and time. Watch it go from Processing to Completed.
Rotate the model, enter its real height and download the GLB.
Generate 3D Model is grayed out (disabled). You haven't added a source image, or your plan or credits don't allow it. Read the message on screen.
My upload was rejected (Only JPG, PNG, and WebP images are allowed / Image size must be less than 10MB). The file isn't JPG, PNG or WebP, or it's over 10 MB.
History says Processing and there's no preview (Preview Unavailable). It's still running. Wait for the state to change.
The model is wrong or distorted. Try a cleaner photo: one object, plain background, fills the frame. Scaling won't fix it.
It says the height isn't ready or unavailable. Wait for the preview to finish loading. If it persists, reopen the History entry and try again.
The model is the wrong size. Check Imperial vs Metric and the decimal point, then re-enter the target height.
I closed the window mid-generation. Closing a draft asks you to confirm discarding the local session. The job itself is listed in History. Reopen AI Image to 3D and find it by thumbnail and time.
Open the Catalog panel, choose AI Image to 3D, upload a JPG, PNG or WebP photo up to 10 MB, and click Generate 3D Model once. When History shows Completed, review the model, enter its real height in Target Height, and download the GLB. Plus or Pro required.
Use one object on a simple or solid-color background, with a clear outline, no text overlays or watermarks, and the whole object visible. The object should fill more than half the frame. A staged scene with several tables, chairs and flowers gives poor results.
Preview Unavailable is not a failure. A job that is still processing has no preview yet, so there is nothing to open. Wait for the state to change to Completed with Preview Ready. Submitting again starts a second job and a second charge.
In Model Transform, check whether the toggle is Imperial (inches) or Metric (centimeters), then type the real height in Target Height. There is no Apply button; the scale factor updates when you leave the field. In the sofa example, 43 inches gave a factor of about 2.166.
Credits are charged when the job starts. When a job ends as failed or cancelled, a refund appears in your credit history as a Feature Refund. A job that completes is not refunded, even if you dislike the model. Confirm unclear adjustments in your credit history.
No. Scaling only resizes the model by the same factor in every direction. It cannot repair a distorted surface or add a detail the generation missed. If the model is wrong, generate again from a cleaner photo: one object, plain background, filling the frame.