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. • 14 items • Updated
Published by TokForge for on-device inference.
TokForge
- Website: https://tokforge.ai
- Discord: https://discord.gg/Acv3CBtfVm
- Google Play: https://play.google.com/store/apps/details?id=dev.tokforge
- iOS TestFlight: https://testflight.apple.com/join/jnufjzRr
Runs on-device in the TokForge app.
TokForge: SD1.5 LCM (MNN)
A Latent Consistency Model (LCM) few-step image generator in MNN format, packaged for fast
on-device CPU image generation in the TokForge Android app (dev.tokforge).
It is a standard Stable Diffusion 1.5 pipeline whose UNet has been LCM-distilled, so it produces a coherent image in just 4-8 steps with guidance-free sampling (CFG = 1.0), no separate negative-prompt pass, which is what makes it fast enough to run on a phone CPU.
Based on
- Base:
Lykon/dreamshaper-7: an SD1.5-architecture checkpoint (CLIP text encoder + UNet + VAE). - LCM adapter:
latent-consistency/lcm-lora-sdv1-5: the LCM-LoRA consistency adapter, fused into the UNet.
The fused pipeline is exported to the MNN diffusion 3-file layout (text encoder / UNet / VAE decoder).
Format
MNN runtime. The bundle ships the standard diffusion files:
| File | Role |
|---|---|
unet.mnn (+ .weight) |
LCM-distilled UNet |
text_encoder.mnn (+ .weight) |
CLIP text encoder |
vae_decoder.mnn (+ .weight) |
VAE decoder |
vocab.json, merges.txt, alphas.txt |
tokenizer + scheduler data |
Usage
This bundle is loaded automatically by the TokForge Android app; it is not a standalone diffusers checkpoint. The app fetches the files, loads them through its MNN diffusion engine, and runs few-step (4-8 step, CFG = 1.0) text-to-image generation on the device CPU.
License & attribution
Released under CreativeML OpenRAIL-M, matching its SD1.5 base.
This model is a derivative of Lykon/dreamshaper-7
with the latent-consistency/lcm-lora-sdv1-5
LCM-LoRA adapter. Please retain this attribution and observe the CreativeML OpenRAIL-M use restrictions.
Credits
- DreamShaper 7 by Lykon (Civitai, Hugging Face). Used under the CreativeML OpenRAIL-M license and the creator's Civitai terms. Lykon's Civitai terms: no selling the model or merges of it.
- LCM-LoRA SD1.5 by Latent Consistency (Hugging Face). Used under the CreativeML OpenRAIL++-M license.
- Stable Diffusion v1.5 by Runway and CompVis (Hugging Face). Used under the CreativeML OpenRAIL-M license.
- Converted to MNN with MNN by Alibaba (Apache-2.0).