> For the complete documentation index, see [llms.txt](https://luw-ai.gitbook.io/luw.ai-studio/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://luw-ai.gitbook.io/luw.ai-studio/luw.ai-ai-tools/photo-to-3d.md). # Photo to 3D Make a 3D model from a photo or a few words in the Create tab, the Items tab, by dropping a photo on your project or by asking Archiagent.ai, then place it, find it again in your 3D library and the Ga The catalog won't always have the exact armchair, lamp or sculpture you're after. Luw\.ai Studio can make it for you: give it a photo of the object, or describe it in a few words, and a 3D model arrives a few minutes later, ready to place. Every model made for you is kept in your Luw\.ai 3D library, so it's available in any project, and it's also filed in the **Gallery** of the project you had open. Already have a model from another program? Upload it as a GLB or GLTF file instead. See [Upload your own 3D model](#upload-your-own-3d-model) below. ## Before you start * **Sign in.** Making models and your 3D library belong to your Luw\.ai account. * **Check your credits.** A model from a photo costs 8 credits on the **Symphony** engine or 3 on **Aria**. A model from words costs 1 more, for the picture it starts from. See [Costs](#costs). ## Make a model in the Create tab The **Create** tab in the left rail, just below **Edit with AI**, is the main place to make a model. Choose **3D model** at the top of the panel, then **From a photo** or **From words**. ### From a photo {% stepper %} {% step %} ### Choose the engine Under **Engine**, **Symphony · 8** is selected by default and gives the higher quality. **Aria · 3** is the lower-cost choice. {% endstep %} {% step %} ### Add the photo Drop a photo onto **Drop, paste or pick a photo of the object**, paste one from your clipboard, or click to choose a file. As the hint says: one object on a plain background works best. {% endstep %} {% step %} ### Keep working The photo uploads, and the panel confirms **Making the model. It comes into the gallery and the corner tray in 2–3 minutes.** The Studio shows the 3D view with the Gallery beside it, and a card in the bottom-left corner follows the progress. {% endstep %} {% step %} ### Place it When the corner card says the model is ready, click **Place in scene**. The model follows your pointer; click where it should stand. You can also add it from the Gallery later. {% endstep %} {% endstepper %}
The Create panel on its 3D model tab with From a photo selected, the Engine choice of Symphony · 8 or Aria · 3, and a dashed drop area reading Drop, paste or pick a photo of the object, One object, plain background works best. Ready in 2–3 minutes.

Create › 3D model, from a photo. The engine buttons show what each one costs.

### From words No photo? Describe the object instead. 3D models are always made from a picture, so the Studio first draws one for you to approve. 1. Choose **From words** and pick the engine. 2. Describe the object under **What to make**, for example "A walnut mid-century armchair with a cream bouclé seat". 3. Click **Show me a picture first · 1 credit**. The picture appears in the panel. 4. Not quite right? **Another picture · 1 credit** draws a new one. 5. When you like it, click **Make it 3D**, which shows the price of the engine you chose (8 or 3 credits). From here it works as with a photo. ## Make a model from the Items tab The **Items** tab has its own entry, the **Photo → 3D** tile at the start of the item grid. It opens **3D models from photos**, which is also where your 3D library lives (see [Your 3D library](#your-3d-library)). 1. Drop a photo onto **Make a new 3D model from a photo**, click it to choose a file, or paste an image. PNG, JPEG and WebP work. 2. The line under it shows the cost. Paying customers also see a **Premium quality** switch: on, the model is made at the higher quality for 8 credits; off, it costs 3. 3. The photo uploads (**Uploading your photo…**), the dialog closes and the corner card takes over. Place the model from there when it's ready. ## Drop a photo straight onto your project You can skip the panels altogether. Drag a photo from your computer onto the 3D view or the 2D plan. While you hold it over the view, a label reads **Drop your land, a floor plan or a photo of furniture**. 1. Let go where the object should stand. 2. The Studio uploads the photo and works out what it shows (**Looking at your image…**). 3. For a picture of an object, the card says **This looks like an object.** Click **Make a 3D model · 3 credits**. Paying customers can tick **Premium quality** next to it, which changes the price to 8 credits. 4. About 2–3 minutes later the model appears by itself where you dropped the photo, fitted into the room there. The corner card then names the room it went into. If the Studio guessed wrong, use the link **It's an object — make a 3D model** below the button. If it can't tell at all, it asks **What is this image?** and you choose **An object to model in 3D**. Press `Esc` or click **Cancel** to forget it. {% hint style="info" %} If the model can't go where you dropped it, for instance because you've opened another project in the meantime, the corner card waits for you and says why. Click **Place in scene** to put it down by hand. {% endhint %} The start card of an empty project takes photos too: **Start from an image** on **How do you want to start?** offers the same **Make a 3D model** choice. That model doesn't place itself; place it from its card when it's ready. ## Ask Archiagent for a model In **Edit with AI**, ask [Archiagent.ai](/luw.ai-studio/luw.ai-ai-tools/archiagent.md) for the object and say where it belongs: * "Make a 3D model of a rattan peacock chair and put it in the corner of the sunroom." * "I need a tall fiddle-leaf fig in a white pot next to the living room window." Archiagent checks the catalog and your 3D library first and only makes what's missing. You can also attach a photo to your message and ask it to model that object, which is closer to the photo and one credit cheaper than a description. With **Ask before spending credits** on (the default), each model waits for your OK on a row in the chat. Pick **Symphony** or **Aria**, each with its price, or **Skip**; a skipped model costs nothing. The row then follows the work: **Picturing it first…** when it starts from words, **Making the 3D model…**, **Placing it in the room…**, and finally which room it was added to. When a request needs several paid steps, Archiagent asks you to approve them all at once as a plan. ## Follow progress in the corner tray Model cards collect in the bottom-left corner. They survive a page reload: a model that was still being made carries on and is picked up again when the page comes back. When several are on their way, they fold into a single **Generating ×***n* card that you can open to see each one. | The card says | Meaning | | ------------------------------------------------------------------------- | ---------------------------------------------------------------------------------------------------- | | **Picturing** *name***…** | A model from words: the picture is being drawn first. | | **Making your 3D model…** (or **Making** *name***…**) with a progress bar | The model is being made. About 2–3 minutes. | | *name* **is ready** and **Place in scene** | Done. Click to put it on your pointer and place it. | | *name* **added to** *room* | A model you dropped, or one Archiagent made, has placed itself. The card closes after a few seconds. | | **You're out of credits** and **Get more credits** | Your balance is too low. The link opens your account page on app.luw\.ai. | | **This image can't be used** | The photo was refused. Try a different one. | | **The generator is busy — try again in a minute** | Start the model again shortly. | The cross on a card hides it. Hiding a card while the model is still being made doesn't cancel anything: the finished model is still saved to your 3D models. ## Find it in the Gallery Every model made while a project is open is also filed in that project's **Gallery**, the view next to **3D** and **2D**, with the photo it came from as its picture. Click it to open it large: * **Add to the scene** puts the model on your pointer to place it. * **Download**, **Star** and **Remove from the project** (which only takes it out of this Gallery) are there too. If Luw\.ai reports that a model failed, its card stays in the Gallery with **Delete & refund**; the credits come back within 24 hours. ## Your 3D library Every model made for you lands in your library, newest first, and it's the same library in every project. 1. Open **Items** → **Photo → 3D**. Your models fill the dialog below the drop area. If a new one is missing, click the refresh button at the top. 2. Click a model. A preview opens with the model turning slowly; drag to look around it and scroll to zoom. 3. Click **Add to scene**. The button shows the progress (**Fetching…**, **Optimizing …%**, **Uploading …%**), then the dialog closes with the model on your pointer. 4. Click in the scene to place it. **Back** or **Cancel** returns to the list without adding anything. An empty library shows **No 3D models yet**. Once placed, a generated model behaves like any other item: move it, turn it, resize it or delete it. See [Placing and arranging](/luw.ai-studio/items-and-paint/placing-and-arranging.md). ## How models are sized and optimized * **Size.** Luw\.ai estimates how wide the real object is, and the model is scaled to that width when you place it. Without an estimate it comes in 80 cm wide. Resize it if the guess is off. * **Optimization.** Every model is simplified before it goes into your project, and its textures are reduced to at most 1024 pixels, so the editor stays smooth even with several in a room. Uploaded models get the same treatment. * **Storage.** The optimized file is stored on Luw\.ai with your project, so it loads wherever you open the project. ## Tips for good photos The advice on the drop areas is the most important one: one object on a plain background. A few more things help: * Show the whole object, with nothing cut off at the edges and nothing standing in front of it. * Use even light. Hard shadows and bright reflections can end up baked into the model. * Choose a background that contrasts with the object, such as a dark chair against a light wall. * Shoot from a natural, slightly raised angle so the shape reads clearly. * A clean product shot is ideal. A busy room photo full of things is not. ## Upload your own 3D model Have a model from SketchUp, Blender, a manufacturer's website or anywhere else? Upload it as a file. Uploading costs no credits. 1. Open **Items** and click the **Upload** tile at the start of the grid. **Upload 3D Model** opens. 2. Drop a file onto **Drop GLB/GLTF file here**, or click to browse. `.glb` and `.gltf` files are accepted. 3. Change the **Model name** if you like. It starts as the file name. 4. Click **Upload & Place**. The file is optimized (**Optimizing…**) and uploaded (**Uploading…**), with a percentage for each step. 5. At **Model ready — click to place!** the dialog closes. Click in the scene to put the model down. Uploaded models keep the size they were made at, so check that your file uses real-world units. If the upload stops at **Upload failed**, the message under it says why; click **Try again** for another go. {% hint style="info" %} If your program can export both, choose `.glb`: it keeps the mesh and its textures together in one file. {% endhint %} Uploaded files belong to the project you're working in. They don't appear in the **3D models from photos** library, which only lists models Luw\.ai made for you, so upload the file again to use it in another project. ## Costs | What | Symphony | Aria | | ----------------------------------------- | -------- | ---- | | Model from a photo | 8 | 3 | | Model from words (picture + model) | 9 | 4 | | Another picture before you make the model | 1 | 1 | | Placing a model from your library again | Free | Free | | Uploading your own GLB or GLTF | Free | Free | In the **Items** tab dialog and the drop card on the canvas, a model costs 3 credits, or 8 with **Premium quality** (paying customers only). Asking Archiagent for a model also uses chat credits for the conversation itself, depending on its Thinking level. See [Credits](/luw.ai-studio/account/credits.md).