README.md

metadata

title: Forge-3D — Image to Game-Ready 3D Assets
emoji: 🧊
colorFrom: indigo
colorTo: purple
sdk: gradio
sdk_version: 6.18.0
app_file: app.py
pinned: true
license: apache-2.0
short_description: Image to 3D models with previews for game assets.

Forge-3D

Image (+text) → 3D brick for game assets. Produces clean models + multi-view previews ready for texturing, rigging, and reverse describe flows.

Part of the 9 open Forge Bricks — composable Gradio Spaces for a complete AI-native game asset creation system.

Capabilities

• Image (or image + text) → game-ready 3D (glb/obj)
• Multi-view preview renders (for VLM describe / consistency in other bricks)
• Health-checked cloud fallbacks + strong local ComfyUI / TripoSR path (excellent on 12GB GPUs)
• Rich manifest.json output with lineage support

Quick Start (Local)

# From forge-bricks root (recommended)
./install.sh
source .venv/bin/activate
cd forge-3d
python app.py

Or with uv directly:

cd forge-3d && uv pip install -r requirements.txt && python app.py

API Endpoint (Daggr / Agents)

• /generate_3d — main entry point

Example:

from daggr import GradioNode
model = GradioNode("jkorstad/forge-3d", api_name="/generate_3d", ...)

HF Space & Best Practices

• Optimized for ZeroGPU (orchestration + client calls; heavy lifting can be local or dedicated).
• Uses adapter + health monitoring pattern (inspired by 3d-creator-suite) for reliability.
• Always returns structured manifests for the rest of the pipeline.

Local First (4070 Ti Class)

Primary recommended path: ComfyUI + TripoSR / similar local models for zero cost and no rate limits. The Gradio app runs identically locally.

Related Bricks

• forge-visuals (bidirectional partner — generate concepts then 3D, or 3D renders → text)
• forge-texture, forge-rig, forge-anim, forge-refiner, etc.

See the main Forge Bricks README and approved plan for full daggr composition examples and ForgeDNA integration.

Open, agent-usable building blocks for creators.