TokForge Image Models
Collection
On-device image generation for TokForge: Hexagon NPU (SD1.5 Fast, SDXL Faithful), CPU/GPU DreamShaper-LCM + 6 styles, RealVisXL photoreal. β’ 13 items β’ Updated
TokForge SD1.5-LCM Styled β Pixel Art
A DreamShaper-7 + LCM SD1.5 MNN bundle (the TokForge SD15_LCM_MNN fast few-step image model)
with the Pixel Art art-style LoRA fused into the UNet at export time. Renders a coherent 512px
image in 4 steps on the in-process MNN diffusion engine β full MNN speed, the style baked in.
9-file MNN bundle: unet.mnn(.weight) (INT8, LCM + style fused), reference SD1.5 f16
text_encoder.mnn(.weight) + vae_decoder.mnn(.weight), vocab.json, merges.txt, alphas.txt.
Provenance & licenses
- Base: Lykon/dreamshaper-7 β CreativeML-OpenRAIL-M
- LCM adapter: latent-consistency/lcm-lora-sdv1-5 β openrail++
- Style LoRA: SedatAl/pixel-art-LoRa β CreativeML-OpenRAIL-M
All components are permissively licensed (no Civitai-NC). Fused with diffusers.fuse_lora()
(fp32 ΞW before INT8 quant); CLIP/VAE reused unchanged from the SD1.5 reference (the style LoRA is
UNet-only).
Inference Providers NEW
This model isn't deployed by any Inference Provider. π Ask for provider support