Image-to-Image
Diffusion Single File
comfyui
novel-view-synthesis
camera-control
fp8
qwen-image-edit
Instructions to use Wherebot101/MetaView-ComfyUI with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Diffusion Single File
How to use Wherebot101/MetaView-ComfyUI with Diffusion Single File:
# No code snippets available yet for this library. # To use this model, check the repository files and the library's documentation. # Want to help? PRs adding snippets are welcome at: # https://github.com/huggingface/huggingface.js
- Notebooks
- Google Colab
- Kaggle
Upload README.md with huggingface_hub
Browse files
README.md
CHANGED
|
@@ -50,17 +50,17 @@ This file goes in `models/diffusion_models/`.
|
|
| 50 |
Example workflow ships with the node pack (`workflows/`). Wiring summary:
|
| 51 |
|
| 52 |
```
|
| 53 |
-
LoadImage โโโฌโโถ MetaView3DConditioning (yaw / pitch / radius, auto_radius) โโถ positive โ
|
| 54 |
-
โโโถ TextEncodeQwenImageEdit (image, no vae)
|
| 55 |
-
|
| 56 |
-
|
| 57 |
```
|
| 58 |
|
| 59 |
-
- **Sampler:**
|
| 60 |
- **Prompt:** the trigger phrase `้ๅคด่ง่ง่ฝฌๅฐๆๅฎไฝ็ฝฎ` ("move the camera to the specified position"). Extra text after it is best-effort โ the model's first-class use is promptless scene mutation.
|
| 61 |
- **Camera:** yaw/pitch in degrees; `auto_radius` derives orbit radius from scene depth (recommended).
|
| 62 |
- Wire the **VAE into MetaView3DConditioning** (not into the text encode node) so the reference latent lands at generation resolution.
|
| 63 |
-
-
|
| 64 |
|
| 65 |
## License & attribution
|
| 66 |
|
|
|
|
| 50 |
Example workflow ships with the node pack (`workflows/`). Wiring summary:
|
| 51 |
|
| 52 |
```
|
| 53 |
+
LoadImage โโโฌโโถ MetaView3DConditioning (yaw / pitch / radius, auto_radius) โโถ positive โโถ SamplerCustom
|
| 54 |
+
โโโถ TextEncodeQwenImageEdit (image, no vae) โโโโโโโโโโโโโโโโโโโโโถ (via conditioning node)
|
| 55 |
+
MetaViewDA3Loader โโถ (da3_model) LATENT from MetaView3DConditioning โโถ SamplerCustom + MetaViewSigmas
|
| 56 |
+
MetaViewModelLoader โโถ MODEL โโถ SamplerCustom (euler, CFG 1.0, sigmas from MetaViewSigmas) โโถ VAEDecode โโถ Save
|
| 57 |
```
|
| 58 |
|
| 59 |
+
- **Sampler:** stock `SamplerCustom` + euler at **CFG 1.0**, sigmas from the pack's `MetaViewSigmas` node (8 steps default). Lightning is fused โ do not add another speed LoRA, and use `MetaViewSigmas` rather than a named scheduler (it reproduces the exact Qwen-Image flow-match schedule the model expects).
|
| 60 |
- **Prompt:** the trigger phrase `้ๅคด่ง่ง่ฝฌๅฐๆๅฎไฝ็ฝฎ` ("move the camera to the specified position"). Extra text after it is best-effort โ the model's first-class use is promptless scene mutation.
|
| 61 |
- **Camera:** yaw/pitch in degrees; `auto_radius` derives orbit radius from scene depth (recommended).
|
| 62 |
- Wire the **VAE into MetaView3DConditioning** (not into the text encode node) so the reference latent lands at generation resolution.
|
| 63 |
+
- Precision note: the loader tags the fp8 layers `full_precision_matrix_mult` (Comfy-Org scaled-fp8 convention) โ weights stay fp8 in RAM/VRAM and dequantize to bf16 for each matmul. No launch flags needed; `--fast fp8_matrix_mult` does not apply to these layers (this model's outputs degrade under fp8 activation matmul, so the loader opts out per-layer).
|
| 64 |
|
| 65 |
## License & attribution
|
| 66 |
|