ComfyUI Node
🌫️ LDM Empty Latent
Creates an empty (zero-initialized) latent tensor for text-to-image generation. The LDM f8 VAE uses 4 latent channels with 8× spatial downsampling (e.g. 256×256 pixels → 32×32 latent). Recommended resolutions: 256×256 (native training resolution). Higher resolutions (512, 1024) are possible but may produce artifacts since the model was trained at 256.
🌫️ LDM Empty Latent
- LATENT
◄width256►
◄height256►
◄batch_size1►
CategoryLDM
Inputs (3)
| Name | Type | Default | Description |
|---|---|---|---|
| width | INT | 25664–1024 | Image width in pixels. Must be a multiple of 8. Native training resolution is 256. |
| height | INT | 25664–1024 | Image height in pixels. Must be a multiple of 8. Native training resolution is 256. |
| batch_size | INT | 11–8 | Number of images to generate in a single pass. |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| LATENT | LDM_LATENT | — |