Preparing model
Unconditional image generator
Image Gen Model Under 50 MB.
Entirely local. Run In Browser
Custom GAN implementations with 47.9 MiB FP16 ONNX models for unconditional 256×256 image generation. It executes inside this page through WebGPU, with CPU/WASM used only as a fallback. No API, cloud GPU, prompt encoder, or server-side inference.
- Model
- Landscapes / 47.9 MiB
- Compute
- WebGPU preferred
- Output
- 9 × 256×256 RGB
- Target latency
- < 100 ms / image*
Connecting to the model repository...
*The first GPU image compiles optimized pipelines and may take longer. After warm-up: under 100 ms per image on supported WebGPU hardware.