Description
Turn an existing Static Mesh into a textured, presentation-ready asset without leaving Unreal Engine.
VellForge is an Editor-only material-authoring and shader-production toolkit for Unreal Engine 5.8 on Windows. Its dedicated VellForge LAB combines local diffusion, UV-aware structure guidance, Unreal mesh baking, PBR map processing, material preview, diagnostics, and asset publication in one workspace.
Choose a curated material preset or build a custom material brief with physical tags such as stone, wood, metal, leather, ceramic, cloth, plastic, rust, patina, and moss. VellForge analyzes the selected mesh and UV layer, generates a surface for that atlas, derives an aligned PBR texture set, creates an Unreal material instance, and applies it to the live preview automatically.
Generation runs locally through the native in-process provider. There is no ComfyUI process, Python worker, cloud API, prompt upload, or automatic model download. Seeds, settings, source images, structure data, and cache identity are retained so an accepted result can be reproduced instead of becoming a one-off experiment.
No API Calls, works on your machine.
No external systems on your localhost.
Runs fully offline on your GPU.
Technical details
- Product type: Unreal Engine C++ Editor plugin
- Supported engine: Unreal Engine 5.8
- Supported platform: Windows 64-bit
- Qualified rendering target: D3D12, Shader Model 6
- Default generation profile: SDXL FinalCut, epsilon prediction
- Default sampler: DPM++ 2M / Karras
- Default working resolution: 1024 × 1024
- Selectable working resolutions: 512, 768, 1024, 1536, and 2048
- Published texture size: 4096 × 4096
- Native execution: in-process; no web service or external inference application
- Blueprint dependency: none; core execution is native C++
### Requirements and important limitations
> CUDA performance disclaimer: VellForge does not guarantee a specific generation time or real-time performance. CUDA acceleration requires a compatible NVIDIA GPU, a suitable graphics driver, and correctly installed native runtime components. Actual performance depends heavily on GPU architecture and compute throughput, available VRAM and memory bandwidth, system RAM, CPU performance, PCIe transfer speed, storage throughput, selected internal resolution, sampling steps, structure guidance, upscaling, and other applications using the GPU. The active full-FP32 checkpoint is larger than the VRAM available on many consumer GPUs, so VellForge can memory-map model data, retain parameters in system memory, and stream bounded work to the GPU; as a result, generation can become CPU-, RAM-, storage-, or transfer-bound even when CUDA is active. The 8 GB VRAM qualification machine is a compatibility reference, not a minimum-performance guarantee, and two GPUs with the same VRAM capacity can perform very differently. When a compatible CUDA backend is unavailable, automatic CPU fallback can be substantially slower. Buyers should evaluate performance on their intended hardware and should not assume the generation times shown or measured on another system.