Downloads · 30 days
0
realvisxl-v5-lightning-unet-onnx
d0gr/realvisxl-v5-lightning-unet-onnx
realvisxl-v5-lightning-unet-onnx is a text-to-image model from d0gr. Use it when you need an image from a text prompt. The card lists the license as openrail++.
UNet-only ONNX export of SG161222/RealVisXLV5.0Lightning for in-browser WebGPU inference (onnxruntime-web) in the Generate AI Images extension.
Access
Public
Updated Aug 17, 2026
Repo size
5.1 GB
Likes
0
Public
Repo makeup
Click a slice to open those files.
.onnx_data_11.9 GB · 37%
At a glance
- Task
- Text-to-Image
- License
- openrail++
- Access
- Public
- Created
- Aug 17, 2026
- Updated
- Aug 17, 2026
- SHA
- 963e32e4
Base models
Model card
From the Hugging Face model README
RealVisXL V5.0 Lightning — UNet ONNX (WebGPU)
UNet-only ONNX export of SG161222/RealVisXL_V5.0_Lightning for in-browser WebGPU inference (onnxruntime-web) in the Generate AI Images extension.
Only the UNet is hosted here: RealVisXL's text encoders and VAE are within fp16 noise of stock SDXL, so the engine shares the copies from d0gr/sdxl-lightning-onnx-webgpu.
Files
| file | role |
|---|---|
unet/model_io32.onnx | graph — fp32 I/O, INT64 timestep, fp16 weights/compute (the profile validated on WebGPU) |
unet/model.onnx | graph — original fp16-I/O self-export profile (kept for reference) |
unet/model.onnx_data, _1, _2 | weight shards (<2 GB each; shared by both graphs) |
I/O: sample float32 [B,4,H,W] · timestep int64 [steps] · encoder_hidden_states
float32 [B,77,2048] · text_embeds float32 [B,1280] · time_ids float32 [B,6] →
out_sample float32 [B,4,H,W].
The checkpoint is Lightning-distilled upstream — run it at 4–6 steps, CFG 1–2,
Euler with trailing spacing. Exported with
tools/export-sdxl-unet.py (LIGHTNING_STEPS=0),
retargeted with tools/retarget-unet-profile.py; one UNet pass agrees with the
torch fp16 checkpoint at cosine 1.000000.
License
CreativeML Open RAIL++-M, inherited from the base model — including its use restrictions (Attachment A). All credit for the weights goes to SG161222; this repo only re-packages them for browser inference.
- Task
- Text-to-Image
- License
- openrail++
- Created
- Aug 17, 2026
- Updated
- Aug 17, 2026