Asset Tap
From prompt, image, or mesh to a rigged bundle.
Your machine, your providers, the whole asset lifecycle.
Asset Tap turns a prompt, an image, or a mesh into a game-ready asset: concept art, a textured GLB, humanoid rigging with animation clips, all in one bundle. Desktop app, CLI, and MCP server. Open source, with generation routed to your chosen provider.
Install
Two things you can install, pick what you need (or both): the CLI for
your terminal, and the desktop app (GUI).
Install the CLI (macOS / Linux)
curl -fsSL https://assettap.dev/install | bash
Installs asset-tap and the atap alias to ~/.local/bin, checksum-verified against the release's SHA256SUMS. Pin a version with ... | bash -s -- v26.8.12; change the destination with ASSET_TAP_INSTALL_DIR=/some/bin. On Windows (PowerShell):
irm https://assettap.dev/install.ps1 | iex
Installs asset-tap.exe and the atap alias to %LOCALAPPDATA%\AssetTap\bin and adds it to your user PATH, checksum-verified. Pin with $env:ASSET_TAP_VERSION = "v26.8.12" first.
Then:
asset-tap auth set fal.ai # or: meshy, tripo
asset-tap "a stylized sci-fi crate"
Install the Desktop App
Download your platform's installer from GitHub Releases.
macOS (Universal: Intel + Apple Silicon)
- Download AssetTap-macos.dmg
- Open the DMG file
- Drag AssetTap to your Applications folder
- Launch from Applications or Spotlight
The app carries its own copy of the CLI internally; for terminal use, run the CLI install above.
Windows
- Download asset-tap-windows-setup.exe
- Run the installer
- Launch from the Start Menu
Linux
.deb (Debian/Ubuntu)
curl -LO https://github.com/nightandwknd/asset-tap/releases/latest/download/asset-tap-linux-amd64.deb
sudo dpkg -i asset-tap-linux-amd64.deb
Installs asset-tap-gui, asset-tap, and the atap alias system-wide.
AppImage (Universal)
curl -LO https://github.com/nightandwknd/asset-tap/releases/latest/download/asset-tap-linux-x86_64.AppImage
chmod +x asset-tap-linux-x86_64.AppImage
./asset-tap-linux-x86_64.AppImage
The AppImage is the GUI only. Pair it with the CLI install above for terminal use.
Build from Source
git clone https://github.com/nightandwknd/asset-tap.git
cd asset-tap
make build
See docs/DEVELOPMENT.md for detailed setup instructions.
Quick Start
1. Get an API Key
Asset Tap ships with pre-configured provider integrations. Choose one or more AI providers that offer text-to-image and image-to-3D capabilities:
Included providers. Pick any one; a single key runs the full pipeline:
You can also add your own providers by creating YAML configuration files (see provider configs in providers/ directory).
2. Launch the Application
Open Asset Tap from your Applications folder, Start Menu, or app launcher. On first launch, you'll be prompted to configure your API key.
3. Generate Your First Model
- Enter a text prompt (e.g., "a cowboy ninja with a leather duster, bandana mask, and dual katanas on the back")
- Select your provider and models
- Click Generate
- Preview your 3D model in the built-in viewer
- Export as GLB
Features
- Built-in 3D Viewer - Preview and inspect models before export
- Multiple AI Models - Choose the best text-to-image and image-to-3D models for your workflow
- Template System - Create and reuse prompt templates
- Image-only mode - Stop after text-to-image when you just want a 2D result
- Reuse past images - Right-click any library item or preview image and pick "Use for Generation" to skip text-to-image on the next run
- Humanoid rigging - Rig a character and bake animation clips (experimental). Free Standard packs install on demand (
asset-tap clip download)
- Library Management - Browse and organize your generated models
- Real-time Progress - Watch generation stages in real-time
CLI Usage
Release archives also install atap, a short alias for asset-tap (a
symlink on macOS/Linux, atap.cmd on Windows). Everything below works with
either name; asset-tap is the canonical one. Source builds don't ship the
alias. Add your own (alias atap=asset-tap).
For automation and scripting:
# Basic generation
asset-tap "a wooden treasure chest"
# Specify provider and models
asset-tap -p fal.ai --image-model fal-ai/nano-banana-2 "a dragon"
# Use an existing image instead of generating one
asset-tap --image "photo.png"
# Stop after image generation, producing an image-only bundle with no 3D model
asset-tap --image-only -y "a wooden treasure chest"
# Copy the finished artifact straight into your project (model.glb, or
# image.png under --image-only). A directory destination is named after
# --name, or the bundle folder.
asset-tap -y --install Assets/Models/crate.glb "a wooden crate"
asset-tap --image-only -y --install sprites/idle.png "a goblin archer idle pose"
asset-tap -y --install Assets/Models/ --name crate "a wooden crate"
# List available providers and models
asset-tap --list-providers
# Auto-confirm the image approval step (only matters if you have approval enabled)
asset-tap -y "a wooden treasure chest"
# Store an API key without launching the GUI (prompts with no echo, or pipe it)
asset-tap auth set fal.ai
echo "$FAL_KEY" | asset-tap auth set fal.ai
asset-tap auth list
# Machine-readable output for tooling: NDJSON events on stdout, one per line
asset-tap --json "a wooden treasure chest"
# Machine-readable model/template catalog (single JSON document)
asset-tap --list-providers --json
asset-tap --list --json
--json mode is a stable wire-format contract for external tools (e.g. the
editor integrations) that drive asset-tap as a subprocess. It emits newline-delimited
JSON events on stdout (all human logs go to stderr), ends with a single
authoritative result event, and uses differentiated exit codes. The full
contract is in docs/CLI_MACHINE_INTERFACE.md.
Driving asset-tap from an AI coding agent (Claude Code, Cursor, Codex, …)?
The CLI is built to be agent-legible: --help, --machine-help, --list --json,
auth list --json, and stable exit codes are the whole interface. Start with
AGENTS.md. Hosts without a shell (Claude Desktop, Cursor) can add
asset-tap mcp as an MCP server. See docs/MCP.md.
See the documentation site for advanced usage.
Output
Generated assets are saved to timestamped directories:
output/
└── 1984-01-24_120000/
├── bundle.json # Metadata (v2: artifacts + pipeline)
├── image.png # AI-generated image
├── model.glb # 3D model (GLB format)
└── textures/ # Extracted textures
Available Models
Text-to-Image
| Model | Provider | Description |
|---|
| Nano Banana | fal.ai | Google Imagen 3-based, fast and affordable |
| Nano Banana 2 | fal.ai | Gemini 3.1 Flash Image, reasoning-guided generation (default) |
| Nano Banana Pro | fal.ai | Premium Imagen 3, higher quality with aspect ratio control |
| FLUX.2 Dev | fal.ai | Open-source FLUX.2 with tunable guidance and steps |
| FLUX.2 Pro | fal.ai | Premium FLUX.2, best quality and zero-config |
| Nano Banana | Meshy | Meshy's standard text-to-image tier (default) |
| Nano Banana 2 | Meshy | Meshy's mid text-to-image tier |
| Nano Banana Pro | Meshy | Meshy's higher-quality text-to-image tier |
| GPT Image 2 | Meshy | The only Meshy model offering 3:2 and 2:3 aspect ratios |
| Seedream v4 | Tripo | ByteDance Seedream 4, balanced quality (default) |
| Seedream v5 | Tripo | ByteDance Seedream 5, stronger prompt following |
| Nano Banana Pro | Tripo | Higher quality with aspect ratio control |
| GPT Image 2 | Tripo | OpenAI GPT Image 2 with a quality tier |
Image-to-3D
| Model | Provider | Description |
|---|
| TRELLIS 2 | fal.ai | Native 3D generative model, fast and versatile (default) |
| Hunyuan3D Pro | fal.ai | Tencent Hunyuan3D v3.1 Pro, high quality 3D generation |
| Meshy v6 | fal.ai | Meshy 6 through fal, pay-per-call billing |
| Meshy v7 | fal.ai | Meshy 7 through fal, pay-per-call billing |
| Meshy 6 Lite | Meshy | Budget tier, lower credit cost |
| Meshy v6 | Meshy | Meshy 6, production-ready 3D with PBR textures |
| Meshy v7 | Meshy | Meshy 7.1, newest generation, high-resolution geometry (default) |
| Smart Topology | Meshy | Meshy T2, clean topology and game-ready face counts |
| Tripo v3.1 | Tripo | Latest H-series model, highest detail (default) |
| Tripo P1 | Tripo | Low-poly with clean topology, 50-20,000 faces |
(default) marks each provider's own default model.
Models are provided by fal.ai, Meshy AI, and Tripo3D. See Provider Documentation for complete details and custom provider setup.
Requirements
- Operating System: macOS 10.15+, Linux (glibc 2.31+), Windows 10+
- AI Provider: API key from fal.ai, Meshy, or Tripo (one is enough)
Documentation
User Guides
Technical Documentation
Troubleshooting
"Provider not found"
- Verify your API key is set correctly
- Check that environment variable matches provider requirements
- Settings → API Keys in the GUI, or
asset-tap auth set <provider> from the CLI
- Run
asset-tap auth list to see which providers have a key and where it's loaded from
Model generation fails
- Check your API key has sufficient credits
- Verify network connection
- Check provider status pages
License
This project is dual-licensed under either of
at your option.
For Developers
See docs/DEVELOPMENT.md for setup, building from source, and contribution workflow.
Support
