Free, open-source 3D sculpting for Linux, macOS and Windows: dense meshes, remeshing, painting, built-in AI (image/text to 3D), Send to Unreal, game-ready LODs, Gaussian splats and OpenUSD.
C++ 3d3d-sculptingcppdear-imguidigital-sculptingfbxgame-developmentgaussian-splattinggltfimage-to-3dmcpopenglopenusdsculptingtext-to-3dunreal-engine
<!doctype html>
<html lang="en">
<head>
<meta charset="utf-8">
<meta name="viewport" content="width=device-width, initial-scale=1">
<title>Image and text to 3D - SculptForge User Guide</title>
<meta name="description" content="SculptForge can turn a picture or a written description into a full 3D model that you can sculpt straight away. This chapter covers the Generate 3D window, the services it can use (free Hugging Face models, your own ComfyUI server, paid fal.ai or Meshy, your own command line, or an offline blockout)">
<meta name="theme-color" content="#111114">
<meta name="color-scheme" content="dark">
<link rel="icon" type="image/png" sizes="32x32" href="../img/favicon-32.png?v=7">
<link rel="icon" type="image/png" sizes="64x64" href="../img/favicon-64.png?v=7">
<link rel="apple-touch-icon" href="../img/apple-touch-icon.png?v=7">
<link rel="canonical" href="https://sculptforge.app/guide/generate-3d.html">
<meta property="og:type" content="article">
<meta property="og:title" content="Image and text to 3D - SculptForge User Guide">
<meta property="og:description" content="SculptForge can turn a picture or a written description into a full 3D model that you can sculpt straight away. This chapter covers the Generate 3D window, the services it can use (free Hugging Face models, your own ComfyUI server, paid fal.ai or Meshy, your own command line, or an offline blockout)">
<meta property="og:image" content="https://sculptforge.app/img/social-preview.png?v=5">
<meta property="og:url" content="https://sculptforge.app/guide/generate-3d.html">
<link rel="stylesheet" href="../style.css?v=19">
<script src="../motion.js?v=12" defer></script>
<script data-goatcounter="https://sculptforge.goatcounter.com/count" async src="https://gc.zgo.at/count.js"></script>
<script src="search-index.js?v=52" defer></script>
</head>
<body class="guide-page" data-prev="retopology.html" data-next="splats.html">
<div class="scroll-progress" aria-hidden="true"></div>
<header class="nav">
<div class="wrap nav-inner">
<a class="brand" href="../index.html"><img class="brand-mark" src="../img/bust-mark-44.png" srcset="../img/bust-mark-88.png 2x, ../img/bust-mark-132.png 3x" width="44" height="44" alt=""><span>SculptForge</span></a>
<nav aria-label="Main">
<a href="../index.html#features">Features</a>
<a href="../index.html#demos">Demos</a>
<a href="../brushes.html">Brushes</a>
<a href="index.html" aria-current="page">Guide</a>
<a class="nav-dl" href="../index.html#download">Download</a>
<a class="gh" href="https://discord.gg/gWrYkEmQhH">Discord</a>
<a class="gh" href="https://github.com/islandmed/SculptForge">GitHub</a>
</nav>
</div>
</header>
<div class="wrap guide-layout">
<div class="guide-side">
<button class="toc-toggle" type="button" aria-expanded="false" aria-controls="guide-toc">
<span>Contents</span><span class="toc-current">Image and text to 3D</span></button>
<div id="guide-toc" class="toc-panel">
<div class="guide-search" role="search">
<input type="search" id="guide-q" placeholder="Search the guide" aria-label="Search the guide" autocomplete="off">
<div class="search-results" id="guide-results" hidden></div>
</div>
<nav class="toc" aria-label="User guide"><a class="toc-home" href="index.html">Contents</a><p class="toc-group">Getting started</p><ul><li><a href="welcome.html">Welcome to SculptForge</a></li><li><a href="first-sculpt.html">Your first sculpt in 10 minutes</a></li><li><a href="whats-new.html">What's new</a></li><li><a href="quick-tour.html">Quick tour for experienced sculptors</a></li><li><a href="installing.html">Install SculptForge on Windows, macOS and Linux</a></li><li><a href="interface.html">The interface</a></li><li><a href="navigation.html">Navigating the view</a></li></ul><p class="toc-group">Sculpting</p><ul><li><a href="sculpting.html">Sculpting and brushes</a></li><li><a href="brush-lab.html">Brush Lab and sharing brushes</a></li><li><a href="brush-packs.html">Brush packs</a></li><li><a href="masking.html">Masking</a></li><li><a href="face-groups.html">Face Groups and visibility</a></li><li><a href="selection-sets.html">Selection sets and morph targets</a></li><li><a href="parts.html">Parts</a></li><li><a href="linked-duplicates.html">Linked Duplicates</a></li><li><a href="forge-spheres.html">Forge Spheres</a></li><li><a href="tube.html">Tube: draw in 3D space</a></li><li><a href="vertex-paint.html">Vertex Paint and materials</a></li><li><a href="texture-painting.html">Texture painting</a></li><li><a href="uv-editor.html">UV Editor</a></li><li><a href="pbr-painting.html">Material painting (PBR)</a></li><li><a href="fill-layers.html">Fill layers, Mesh Maps and Material Recipes</a></li><li><a href="layers.html">Sculpt layers and blend shapes</a></li><li><a href="facial-shapes.html">Facial shapes and live face</a></li><li><a href="subdivision-levels.html">Subdivision levels</a></li><li><a href="micro-levels.html">Micro levels: hundreds of millions of polygons</a></li><li><a href="dynamic-topology.html">Dynamic topology</a></li><li><a href="timeline.html">Sculpt Timeline and timelapse</a></li><li><a href="gizmo.html">Moving, rotating and posing (Gizmo)</a></li><li><a href="geometry.html">Subdivide, remesh, decimate and LODs</a></li><li><a href="cut.html">Cut: line, lasso, rectangle and polyline</a></li><li><a href="retopology.html">Quad retopology and detail projection</a></li></ul><p class="toc-group">AI and capture</p><ul><li><a href="generate-3d.html" class="current" aria-current="page">Image and text to 3D</a><ul class="toc-sections"><li><a href="#how-it-works">How it works</a></li><li><a href="#open-the-generate-3d-window">Open the Generate 3D window</a></li><li><a href="#generate-from-an-image">Generate from an image</a></li><li><a href="#generate-from-text">Generate from text</a></li><li><a href="#services">Services</a></li><li><a href="#save-your-settings">Save your settings</a></li><li><a href="#quick-shapes-from-an-image-offline">Quick shapes from an image (offline)</a></li><li><a href="#after-generating">After generating</a></li><li><a href="#from-scripts-and-agents">From scripts and agents</a></li><li><a href="#privacy-and-model-licences">Privacy and model licences</a></li><li><a href="#see-also">See also</a></li></ul></li><li><a href="splats.html">Gaussian splats</a></li><li><a href="scan-to-splat.html">Scan an object into a splat</a></li><li><a href="scan-to-sculptable.html">From scan to sculptable mesh</a></li><li><a href="phone-capture.html">Phone capture and AR</a></li></ul><p class="toc-group">Files and pipelines</p><ul><li><a href="baking.html">Baking normal and displacement maps</a></li><li><a href="exporting.html">Exporting, Unreal and Bridge</a></li><li><a href="export-presets.html">Export presets</a></li><li><a href="blender.html">Working with Blender</a></li><li><a href="maya.html">Working with Maya</a></li><li><a href="projects.html">Projects, autosave and snapshots</a></li><li><a href="scripting.html">Scripting and automation</a></li><li><a href="packs-and-macros.html">Packs and macros</a></li><li><a href="studios.html">SculptForge for studios</a></li><li><a href="security.html">Security overview for IT reviews</a></li></ul><p class="toc-group">Reference</p><ul><li><a href="shortcuts.html">Keyboard and mouse</a></li><li><a href="performance.html">Performance and big meshes</a></li><li><a href="troubleshooting.html">Troubleshooting: tablets, display, crashes and FAQ</a></li><li><a href="glossary.html">Glossary</a></li><li><a href="discord.html">Show SculptForge on Discord</a></li></ul></nav>
</div>
</div>
<main class="guide-main" id="top">
<article class="guide-article">
<h1 id="image-and-text-to-3d">Image and text to 3D</h1>
<p>SculptForge can turn a picture or a written description into a full 3D model that you can sculpt straight away. This chapter covers the <strong>Generate 3D</strong> window, the services it can use (free Hugging Face models, your own ComfyUI server, paid fal.ai or Meshy, your own command line, or an offline blockout) and the offline image tools: relief, inflate and stamp.</p>
<figure class="shot"><a href="images/generate-3d.jpg" class="zoom"><img src="images/generate-3d.jpg" alt="The Generate 3D window with a sketch loaded and the Hugging Face service selected" width="316" height="728" loading="lazy" decoding="async"></a><figcaption>The Generate 3D window with a sketch loaded and the Hugging Face service selected</figcaption></figure>
<h2 id="how-it-works">How it works<a class="anchor" href="#how-it-works" aria-label="Link to this section">#</a></h2>
<ol>
<li>
You give SculptForge an image (a sketch, concept art, a photo of a figurine) or a short description.
</li>
<li>
The service you picked reconstructs a complete mesh: front, back, limbs and all.
</li>
<li>
SculptForge opens the result as a new scene and, unless you choose otherwise, <strong>Quad Remeshes</strong> it into clean quads with one subdivision level that holds the detail, ready to sculpt and subdivide.
</li>
</ol>
<p>Nothing leaves your computer until you choose a hosted service and press <strong>Generate 3D model</strong>. For exactly what is sent where, see <a href="#privacy-and-model-licences">AI services and privacy</a> below.</p>
<h2 id="open-the-generate-3d-window">Open the Generate 3D window<a class="anchor" href="#open-the-generate-3d-window" aria-label="Link to this section">#</a></h2>
<p>Any of these opens it:</p>
<ul>
<li>
Drop an image file (PNG, JPG, BMP, TGA, PSD or GIF) onto the SculptForge window. The image is loaded as the picture for the <strong>From image</strong> tab and the window opens.
</li>
<li>
Open <strong>Mesh > Generate 3D</strong> and click <strong>Generate 3D</strong>.
</li>
<li>
Press <kbd>F3</kbd>, type <code>generate</code> and press <kbd>Enter</kbd>.
</li>
</ul>
<p>The window has two tabs, <strong>From image</strong> and <strong>From text</strong>.</p>
<aside class="callout callout-warning"><p><strong>Warning:</strong> A generated model replaces the current scene without asking, and arrives as a new, untitled scene. Save your work (<kbd>Ctrl</kbd>+<kbd>S</kbd>) before you generate.</p></aside>
<h2 id="generate-from-an-image">Generate from an image<a class="anchor" href="#generate-from-an-image" aria-label="Link to this section">#</a></h2>
<ol>
<li>
On the <strong>From image</strong> tab, click <strong>Choose image...</strong> (or drop a file onto the window). A preview and the image size appear.
</li>
<li>
Pick a service in the drop-down under <strong>Full 3D model (AI reconstruction)</strong>. The default is <strong>Hugging Face</strong>
<ul>
<li>
free models<strong>.</strong>
</li>
</ul>
</li>
</ol>
<ol start="3">
<li>
Set the service's options (model, token or key, see below).
</li>
<li>
Choose <strong>Remesh</strong> and its <strong>Resolution</strong> (96 to 400, default 256; higher keeps more detail):
<ul>
<li>
<strong>Quad Remesh</strong> (the default): clean, all-quad topology of about resolution² / 2 quads (32768 at 256), then one subdivision level projected onto the model so the detail survives. Generated meshes often have holes and overlapping pieces, so a voxel pass cleans the mesh up first. Step down to the quad cage with <kbd>Shift</kbd>+<kbd>D</kbd>.
</li>
<li>
<strong>Voxel remesh</strong>: one even, watertight triangle surface. Faster (about a second).
</li>
<li>
<strong>Off (raw mesh)</strong>: the model exactly as the service made it.
</li>
</ul>
</li>
</ol>
<ol start="5">
<li>
Leave <strong>Remove flat background</strong> on (the default) to drop the flat plane a picture's background often becomes: a thin wall behind or a ground under the subject that covers most of the model's bounds. A relief or a coin on its own is kept. The status line says when a plane was removed; Undo brings it back (with Remesh on, the first Undo steps back to the raw mesh, the second brings the plane back).
</li>
<li>
Click <strong>Generate 3D model</strong>. A progress bar shows the queue position and progress; <strong>Cancel</strong> stops the job.
</li>
</ol>
<p>When it finishes, the status line reports the size before and after remeshing, for example "Generated model imported (47214 tris, remeshed to 133540 quads (2 subdivision levels)) - ready to sculpt". Quad Remesh takes about 10 to 25 seconds after the model arrives (a 47,000-triangle Hunyuan3D head took 22 s on a busy 10-core computer); the progress bar shows it and the window stays usable.</p>
<h3 id="pictures-that-work-well">Pictures that work well<a class="anchor" href="#pictures-that-work-well" aria-label="Link to this section">#</a></h3>
<ul>
<li>
One character or object per image.
</li>
<li>
A plain or transparent background. SculptForge crops the subject out of a transparent background, centres it on a square and scales it to at most 1024 pixels before sending it.
</li>
<li>
The whole subject in frame, seen from the front or a three-quarter view.
</li>
<li>
A clean concept turnaround pose works better than a busy illustration with scenery.
</li>
</ul>
<h2 id="generate-from-text">Generate from text<a class="anchor" href="#generate-from-text" aria-label="Link to this section">#</a></h2>
<ol>
<li>
On the <strong>From text</strong> tab, type one object into the box, for example "a stylised dragon head with curled horns" or "a cartoon fox sitting".
</li>
<li>
Pick <strong>Hugging Face - free models</strong>, <strong>fal.ai</strong>, <strong>Meshy</strong> or <strong>ComfyUI - your server</strong>. Text to 3D does not work with <strong>My generator command</strong> or <strong>Offline blockout</strong>.
</li>
<li>
Click <strong>Generate 3D model</strong>.
</li>
</ol>
<p>Which model runs depends on the service:</p>
<div class="table-wrap"><table><thead><tr><th>Service</th><th>Text-to-3D model</th></tr></thead><tbody><tr><td>Hugging Face</td><td>Hunyuan3D 2 mini Turbo (Tencent), or Hunyuan3D 2 if you chose it in <strong>Model</strong>. Any other choice uses Turbo</td></tr><tr><td>fal.ai</td><td>Rodin 2.5 (Hyper3D) when it is the chosen <strong>Model</strong> (Rodin 2.5 or Rodin 2.5 Fast), otherwise Hunyuan3D 3.1 Rapid (Tencent)</td></tr><tr><td>Meshy</td><td>Meshy's own text-to-3D model</td></tr><tr><td>Hyper3D Rodin (direct)</td><td>Rodin Gen-2.5 at the chosen <strong>Tier</strong></td></tr><tr><td>ComfyUI</td><td>The text workflow you choose (the bundled <code>text-to-3d</code> uses FLUX.1 schnell and Hunyuan3D 2.1)</td></tr></tbody></table></div>
<h2 id="services">Services<a class="anchor" href="#services" aria-label="Link to this section">#</a></h2>
<div class="table-wrap"><table><thead><tr><th>Service</th><th>Cost and setup</th><th>Good for</th></tr></thead><tbody><tr><td><strong>Hugging Face - free models</strong></td><td>Free, no account. A free token raises the daily allowance</td><td>Trying things out, personal work</td></tr><tr><td><strong>fal.ai - paid, your key</strong></td><td>Paid per generation, needs a fal.ai key</td><td>Reliable capacity, newer models</td></tr><tr><td><strong>Meshy - paid, your key</strong></td><td>Paid, needs a Meshy key</td><td>Meshy's own models</td></tr><tr><td><strong>ComfyUI - your server</strong></td><td>Free, runs on your own GPU or a studio server</td><td>Privacy, custom workflows, studios</td></tr><tr><td><strong>My generator command</strong></td><td>Free, any program you run locally</td><td>Your own model or a studio service</td></tr><tr><td><strong>Offline blockout (inflate silhouette)</strong></td><td>Free and instant, no network</td><td>Simple shapes, a quick start</td></tr></tbody></table></div>
<h3 id="hugging-face-free">Hugging Face (free)<a class="anchor" href="#hugging-face-free" aria-label="Link to this section">#</a></h3>
<p>This runs each model author's own public demo on Hugging Face's shared ZeroGPU hardware. Pick a <strong>Model</strong>:</p>
<div class="table-wrap"><table><thead><tr><th>Model</th><th>Notes</th></tr></thead><tbody><tr><td>Hunyuan3D 2.1 (Tencent)</td><td>The default. Detailed shapes, about 20 seconds, no token needed</td></tr><tr><td>TRELLIS.2 (Microsoft)</td><td>Best detail, MIT licence. Needs a free Hugging Face token</td></tr><tr><td>Hunyuan3D 2 mini Turbo (Tencent)</td><td>The fastest, about 6 seconds</td></tr><tr><td>Hunyuan3D 2 (Tencent)</td><td></td></tr><tr><td>Stable Fast 3D (Stability AI)</td><td>Quick and simple, removes the background itself</td></tr></tbody></table></div>
<p>If you never picked a model, SculptForge uses the default, so updates that change it reach you; a model you pick (and save) is kept. If you pick TRELLIS.2 without a token, Hunyuan3D 2.1 runs instead and the status line says "TRELLIS.2 needs a free Hugging Face token (Edit > Preferences > AI); used Hunyuan3D 2.1 instead."</p>
<p>The Hunyuan3D models' licence doesn't cover use in the European Union, the United Kingdom or South Korea. The first time a Hunyuan3D model runs, a notice says so, and it stays in the <strong>Model</strong> picker's tooltip. See <a href="https://github.com/islandmed/SculptForge/blob/main/docs/ai-services.md" rel="noopener">AI services</a> for each model's licence.</p>
<p>GPU time is rationed per person per day, there can be a queue at busy times, and a demo can be paused or changed by its owner. To raise the allowance, create a free token with read access at huggingface.co/settings/tokens and paste it into <strong>HF token</strong> or <strong>Edit > Preferences > AI</strong> (or set the <code>HF_TOKEN</code> environment variable). When the allowance runs out you see "Hugging Face's free GPU time is used up for now"; add a token, wait for it to refill, or pick another model.</p>
<aside class="callout callout-warning"><p><strong>Warning:</strong> The Hunyuan3D models are under the Tencent Hunyuan Community License, which does not cover use in the European Union, the United Kingdom or South Korea. If you work there, choose TRELLIS.2 (free token), Stable Fast 3D or a paid service. Each model has its own licence; read it before using results commercially.</p></aside>
<h3 id="falai-meshy-and-hyper3d-paid">fal.ai, Meshy and Hyper3D (paid)<a class="anchor" href="#falai-meshy-and-hyper3d-paid" aria-label="Link to this section">#</a></h3>
<ul>
<li>
<strong>fal.ai</strong>: choose a <strong>Model</strong> (TRELLIS.2, Hunyuan3D 3.1 Pro, Hunyuan3D 3.1 Rapid, Meshy 6, Rodin 2.5 or Rodin 2.5 Fast) or type any fal model id into <strong>Model id</strong>. Rodin 2.5 (Hyper3D) works from an image or a description; SculptForge asks it for the shape only (GLB, no materials). Rodin 2.5 Fast is cheaper and low poly (up to about 20K triangles), fine as a blockout before remeshing. Paste your key from fal.ai/dashboard/keys into <strong>API key</strong>, or set <code>FAL_KEY</code>.
</li>
<li>
<strong>Meshy</strong>: paste your key from meshy.ai (API settings) into <strong>API key</strong>, or set <code>MESHY_API_KEY</code>.
</li>
<li>
<strong>Hyper3D Rodin (direct)</strong>: Rodin Gen-2.5 on Hyper3D's own service, from an image or a description. Choose a <strong>Tier</strong> (Extreme Low is the fastest; Extreme High costs more credits) and paste your key from the API dashboard at hyper3d.ai into <strong>API key</strong>, then <strong>Save settings</strong>: the key goes into the system keychain. It stays off until you add a key. Your image or description goes to Hyper3D (api.hyper3d.com) and nowhere else, and SculptForge asks for the shape only (GLB, no materials). Basic API access (Creator plans) covers Extreme Low and Low; the other tiers need Full access.
</li>
</ul>
<p>Generations are billed to your account with that service. Errors such as "The API key was rejected", "Out of credits on this account" or "Rate limited" come from the service.</p>
<h3 id="comfyui-your-own-server">ComfyUI (your own server)<a class="anchor" href="#comfyui-your-own-server" aria-label="Link to this section">#</a></h3>
<p>Runs any ComfyUI workflow on your own computer or a studio GPU server, with whatever models it has. Nothing is sent anywhere else.</p>
<ol>
<li>
Start ComfyUI. It listens on <code>http://127.0.0.1:8188</code> by default. If SculptForge runs on another computer, start ComfyUI with <code>--listen</code> (only on a network you trust: ComfyUI has no login).
</li>
<li>
Install the models your workflow needs. The bundled Hunyuan3D 2.1 workflow needs <code>hunyuan_3d_v2.1.safetensors</code> in <code>ComfyUI/models/checkpoints</code>.
</li>
<li>
In <strong>Generate 3D</strong>, pick <strong>ComfyUI - your server</strong>, enter the <strong>Server</strong> address and click <strong>Test connection</strong>. A good answer looks like "Connected: ComfyUI 0.42.0 - NVIDIA GeForce RTX 4090 (24.0 GB)".
</li>
<li>
Choose a <strong>Workflow</strong>: a bundled one, or <strong>Other file (API format)...</strong> for your own.
</li>
<li>
Click <strong>Generate 3D model</strong>. The progress bar shows your place in ComfyUI's queue.
</li>
</ol>
<p>Bundled workflows:</p>
<div class="table-wrap"><table><thead><tr><th>Workflow</th><th>Tab</th><th>Needs</th></tr></thead><tbody><tr><td><code>hunyuan3d-image-to-3d</code></td><td>From image</td><td>The Hunyuan3D 2.1 checkpoint (7.4 GB) and a GPU with about 10 GB of memory</td></tr><tr><td><code>text-to-3d</code></td><td>From text</td><td>FLUX.1 schnell, BiRefNet and Hunyuan3D 2.1 (about 25 GB of downloads) and a GPU</td></tr><tr><td><code>sculptforge-inflate</code></td><td>From image</td><td>The ComfyUI-SculptForge node pack; no GPU. A quick blockout to check your setup</td></tr></tbody></table></div>
<p>Your own workflows must be saved in ComfyUI's <strong>API format</strong> (in ComfyUI: <strong>Workflow > Export (API)</strong>). To tell SculptForge where the inputs go, give nodes these titles:</p>
<div class="table-wrap"><table><thead><tr><th>Node title</th><th>What SculptForge does</th></tr></thead><tbody><tr><td><code>sf_image</code></td><td>Puts the uploaded picture into this Load Image node. Without one, the first Load Image node is used</td></tr><tr><td><code>sf_prompt</code></td><td>Puts your description into this node's <code>text</code>, <code>prompt</code> or <code>string</code> input. Without one, the first CLIP Text Encode not titled "negative" is used</td></tr><tr><td><code>sf_seed</code></td><td>Sets the seed. Inputs named <code>seed</code> or <code>noise_seed</code> get it too</td></tr></tbody></table></div>
<p>SculptForge imports the first mesh file the workflow saves (Save GLB, Preview 3D and most 3D node packs work). For error messages and their fixes, see <a href="troubleshooting.html#comfyui-errors">Troubleshooting</a>.</p>
<h3 id="my-generator-command">My generator command<a class="anchor" href="#my-generator-command" aria-label="Link to this section">#</a></h3>
<p>Runs any command line on your computer: a local open-source model, a studio service, or a script that calls a cloud API. Type the command into the box. It must contain <code>{input}</code> (replaced with the image path) and <code>{output}</code> (replaced with the path to write, ending in <code>.glb</code>). Both are quoted for you, so don't add quotes. The command must write a GLB file to <code>{output}</code> and exit with status 0.</p>
<p>The box starts with a command that runs SculptForge's own silhouette inflation, which is a quick way to check that the pipeline works:</p>
<div class="code-block"><pre class="code"><code class="language-text">sculptforge-cli --set input={input} --set output={output} -c 'image_inflate path=$input subdivisions=2; export_glb path=$output'</code></pre><button class="copy" type="button" aria-label="Copy">Copy</button></div>
<p>A wrapper for your own model can be as small as this:</p>
<div class="code-block"><pre class="code"><code class="language-python"># my_generator.py {input} {output}
import sys
image_path, output_path = sys.argv[1], sys.argv[2]
# load your model, run it on image_path, write a .glb to output_path</code></pre><button class="copy" type="button" aria-label="Copy">Copy</button></div>
<p>Keep API keys in environment variables that your script reads, not in the command line.</p>
<h3 id="offline-blockout">Offline blockout<a class="anchor" href="#offline-blockout" aria-label="Link to this section">#</a></h3>
<p><strong>Offline blockout (inflate silhouette)</strong> inflates the outline of the picture into a rounded, closed mesh. It is instant and needs no network, but it is a puffy blockout, not a reconstruction: fine for simple shapes, not for characters.</p>
<h2 id="save-your-settings">Save your settings<a class="anchor" href="#save-your-settings" aria-label="Link to this section">#</a></h2>
<p><strong>Save settings</strong> (the small button under <strong>Generate 3D model</strong>) stores the service, model, keys, token, ComfyUI address and generator command in your <code>ai.json</code> settings file, so they are set the next time you start SculptForge. Keys in <code>ai.json</code> are readable only by your user account. See <a href="troubleshooting.html#where-settings-live">where settings live</a>.</p>
<h2 id="quick-shapes-from-an-image-offline">Quick shapes from an image (offline)<a class="anchor" href="#quick-shapes-from-an-image-offline" aria-label="Link to this section">#</a></h2>
<p>The <strong>From image</strong> tab also has three offline tools under <strong>Quick shapes (offline)</strong>. They need no network and run instantly. The same tools are in <strong>Mesh > Generate 3D</strong> as <strong>Relief</strong>, <strong>Inflate</strong> and <strong>Stamp</strong>.</p>
<div class="table-wrap"><table><thead><tr><th>Tool</th><th>What it makes</th><th>Settings</th></tr></thead><tbody><tr><td><strong>Silhouette blockout</strong> > <strong>Inflate</strong></td><td>A rounded, closed base mesh from the outline. Replaces the scene</td><td><strong>Thickness</strong> (0.2 to 2, lower is flatter), <strong>Detail</strong> (64 to 384)</td></tr><tr><td><strong>Relief</strong> > <strong>Make relief</strong></td><td>A height-field relief: light areas rise. For logos, coins, emblems, height maps and texture photos. Replaces the scene</td><td><strong>Depth</strong> (0.01 to 0.5 of the width), <strong>Detail</strong> (64 to 1024), <strong>Dark is high</strong></td></tr><tr><td><strong>Stamp onto model</strong> > <strong>Stamp from this view</strong></td><td>Projects the image onto the current model from the current view as surface detail, into the active sculpt layer</td><td><strong>Depth</strong> (0.001 to 0.1 of the model size), <strong>Dark is high</strong></td></tr></tbody></table></div>
<aside class="callout callout-tip"><p><strong>Tip:</strong> Stamp writes into the active sculpt layer, so add a layer first (<strong>Tool > Layers</strong>) if you want to dial the detail in and out afterwards. See <a href="layers.html">Layers and blend shapes</a>.</p></aside>
<h2 id="after-generating">After generating<a class="anchor" href="#after-generating" aria-label="Link to this section">#</a></h2>
<p>Generated meshes are a starting point. A typical next pass:</p>
<ol>
<li>
Check the size of the remeshed result in the bar along the top (the triangle count). If the detail is too soft, divide once more (<kbd>Ctrl</kbd>+<kbd>D</kbd>), or generate again with a higher <strong>Resolution</strong>. <strong>Tool > Remesh</strong> has Quad Remesh (first) and Voxel Remesh for another pass.
</li>
<li>
Turn on <strong>X</strong> symmetry if the subject is symmetrical, and sculpt the forms with Clay Block, Move and Sharp Crease (<a href="sculpting.html">Sculpting and brushes</a>).
</li>
<li>
Save a project (<kbd>Ctrl</kbd>+<kbd>S</kbd>): generated scenes start untitled.
</li>
</ol>
<h2 id="from-scripts-and-agents">From scripts and agents<a class="anchor" href="#from-scripts-and-agents" aria-label="Link to this section">#</a></h2>
<p>The same services are available as commands, in the console, <code>sculptforge-cli</code>, Python and MCP agents:</p>
<div class="code-block"><pre class="code"><code class="language-sh"># free, no token (Hunyuan3D 2.1 on Hugging Face, the default)
sculptforge-cli -c 'image_to_3d path=hero.png; export_glb path=hero.glb'
# TRELLIS.2, best detail, with a free Hugging Face token (without one, Hunyuan3D 2.1 runs and says so)
HF_TOKEN=hf_... sculptforge-cli -c 'image_to_3d path=hero.png model=trellis-2; export_glb path=hero.glb'
# keep a flat background plane the service made (removed by default; remove_flat_background does it later)
sculptforge-cli -c 'image_to_3d path=hero.png remove_background=false; export_glb path=hero.glb'
# voxel remesh instead of the default Quad Remesh
sculptforge-cli -c 'image_to_3d path=hero.png remesh=voxel; export_glb path=hero.glb'
# fal.ai, keeping the raw (not remeshed) result
FAL_KEY=... sculptforge-cli -c 'image_to_3d path=hero.png service=fal model=fal-ai/trellis-2 remesh=false; export_glb path=hero.glb'
# text to 3D on your own ComfyUI, repeatable with a seed
sculptforge-cli -c 'text_to_3d prompt="a carved stone owl" service=comfyui url=http://gpu-box:8188 seed=42; export_usd path=owl.usdz'
# is the ComfyUI server up?
sculptforge-cli -c 'comfyui_check'
# offline tools
sculptforge-cli -c 'image_relief path=logo.png depth=0.08; export_stl path=logo.stl'
sculptforge-cli -c 'image_inflate path=sketch.png thickness=0.7; save_project path=blockout.sfproj'</code></pre><button class="copy" type="button" aria-label="Copy">Copy</button></div>
<div class="table-wrap"><table><thead><tr><th>Command</th><th>Key parameters</th></tr></thead><tbody><tr><td><code>image_to_3d</code></td><td><code>path</code> (required), <code>service</code> (<code>huggingface</code>, <code>fal</code>, <code>meshy</code>, <code>comfyui</code>, <code>command</code>, <code>inflate</code>), <code>model</code>, <code>workflow</code>, <code>url</code>, <code>seed</code>, <code>remesh</code> (<code>quad</code>, the default; <code>voxel</code>; <code>false</code>), <code>resolution</code> (64 to 512, default 256), <code>levels</code> (quad: 0 to 2 subdivision levels, default 1)</td></tr><tr><td><code>text_to_3d</code></td><td><code>prompt</code> (required), <code>service</code> (<code>huggingface</code>, <code>fal</code>, <code>meshy</code>, <code>comfyui</code>), <code>model</code>, <code>workflow</code>, <code>url</code>, <code>seed</code>, <code>remesh</code>, <code>resolution</code>, <code>levels</code></td></tr><tr><td><code>image_relief</code></td><td><code>path</code> (required), <code>depth</code> (default 0.12), <code>resolution</code> (default 384), <code>invert</code></td></tr><tr><td><code>image_inflate</code></td><td><code>path</code> (required), <code>thickness</code> (default 1), <code>resolution</code> (default 160), <code>subdivisions</code> (default 1)</td></tr><tr><td><code>image_project</code></td><td><code>path</code> (required), <code>depth</code> (default 0.03), <code>invert</code>. Projects from the front (+Z)</td></tr><tr><td><code>comfyui_check</code></td><td><code>url</code></td></tr></tbody></table></div>
<p>Without <code>service=</code>, commands use the settings you saved in the app. See <a href="scripting.html">Scripting and automation</a>.</p>
<h2 id="privacy-and-model-licences">Privacy and model licences<a class="anchor" href="#privacy-and-model-licences" aria-label="Link to this section">#</a></h2>
<ul>
<li>
Your art is never uploaded or used for training. AI features run only when you use them and can be turned off. Off (<strong>Edit > Preferences > AI</strong>, or <code>SF_AI=off</code>) hides Generate 3D; Relief, Inflate and Stamp stay. See <a href="interface.html#turning-ai-off">Turning AI off</a>.
</li>
<li>
SculptForge has no servers of its own and no telemetry. A hosted service receives only the prepared image (or your text) and your key or token, if any.
</li>
<li>
The free Hugging Face option runs the model authors' own demos; SculptForge does not host or bundle any model.
</li>
<li>
Each model has its own licence: TRELLIS.2 is MIT; Hunyuan3D models use the Tencent Hunyuan Community License (not for use in the EU, UK or South Korea); Stable Fast 3D uses the Stability AI Community License (free under US $1M annual revenue). Read the licence before commercial use.
</li>
<li>
With ComfyUI, the picture goes only to the server address you entered.
</li>
</ul>
<p>The full table of what is sent where is in <code>docs/ai-services.md</code> in the SculptForge repository.</p>
<h2 id="see-also">See also<a class="anchor" href="#see-also" aria-label="Link to this section">#</a></h2>
<ul>
<li>
<a href="sculpting.html">Sculpting and brushes</a>: refine the generated model
</li>
<li>
<a href="geometry.html">Subdivide, Remesh, Decimate and LODs</a>: Quad Remesh and voxel remesh in detail
</li>
<li>
<a href="splats.html">Gaussian splats</a>: start from a phone capture instead of a picture
</li>
<li>
<a href="scripting.html">Scripting and automation</a>: batch generation from the command line
</li>
<li>
<a href="troubleshooting.html#ai-services">Troubleshooting</a>: quota and connection messages
</li>
</ul>
</article>
<nav class="pager" aria-label="Previous and next chapter"><a class="prev" rel="prev" href="retopology.html"><span>Previous</span>Quad retopology and detail projection</a><a class="next" rel="next" href="splats.html"><span>Next</span>Gaussian splats</a></nav>
<p class="guide-meta fine">Also in the app: <strong>Help > User Guide</strong> (<kbd>Shift</kbd>+<kbd>F1</kbd>).
Found a mistake? <a href="https://github.com/islandmed/SculptForge/edit/main/docs/guide/generate-3d.md">Edit this page on GitHub</a>.</p>
</main>
<aside class="on-page" aria-label="On this page"><p>On this page</p><ul><li><a href="#how-it-works">How it works</a></li><li><a href="#open-the-generate-3d-window">Open the Generate 3D window</a></li><li><a href="#generate-from-an-image">Generate from an image</a></li><li><a href="#generate-from-text">Generate from text</a></li><li><a href="#services">Services</a></li><li><a href="#save-your-settings">Save your settings</a></li><li><a href="#quick-shapes-from-an-image-offline">Quick shapes from an image (offline)</a></li><li><a href="#after-generating">After generating</a></li><li><a href="#from-scripts-and-agents">From scripts and agents</a></li><li><a href="#privacy-and-model-licences">Privacy and model licences</a></li><li><a href="#see-also">See also</a></li></ul></aside>
</div>
<footer class="footer">
<div class="wrap footer-grid">
<div>
<a class="brand" href="../index.html"><img class="brand-mark" src="../img/bust-mark-44.png" srcset="../img/bust-mark-88.png 2x, ../img/bust-mark-132.png 3x" width="32" height="32" alt=""><span>SculptForge</span></a>
<p class="fine">Copyright © 2026 The SculptForge contributors.<br>The SculptForge User Guide is part of
SculptForge, licensed under the <a href="https://github.com/islandmed/SculptForge/blob/main/LICENSE">GNU GPL v3.0 or later</a>.</p>
</div>
<div class="links">
<a href="../performance/">Performance</a>
<a href="index.html">User guide</a>
<a href="../supporters.html">Members</a>
<a href="https://github.com/islandmed/SculptForge">Source code</a>
<a href="https://discord.gg/gWrYkEmQhH">Community Discord</a>
<a href="https://github.com/islandmed/SculptForge/issues">Report a problem</a>
<a href="https://github.com/islandmed/SculptForge/blob/main/docs/ai-services.md">AI services & privacy</a>
</div>
<p class="fine trademarks">SculptForge is an independent project. Product names are trademarks of their respective owners.</p>
</div>
</footer>
</body>
</html>