43 lines
2.6 KiB
Markdown
43 lines
2.6 KiB
Markdown
## TT Latent Factory *TT_LatentFactoryNode*
|
||
|
||
*This description was generated with AI assistance. If you spot any errors, please create an issue on GitHub.*
|
||
|
||
Creates a latent tensor and noise for use in sampling pipelines with support for FLUX and SDXL architectures.
|
||
|
||
### Inputs
|
||
|
||
| Parameter | Type | Required | Description |
|
||
|-------------------|-------|----------|-----------------------------------------------------------------|
|
||
| `seed` | INT | Yes | Seed for latent tensor generation |
|
||
| `noise_seed` | INT | Yes | Seed for noise generation (increments after each run) |
|
||
| `aspect_ratio` | COMBO | Yes | Aspect ratio of the latent (1:1, 4:3, 3:2, 16:9, 21:9) |
|
||
| `megapixels` | COMBO | Yes | Target resolution in megapixels |
|
||
| `orientation` | COMBO | Yes | Landscape or portrait orientation |
|
||
| `model_type` | COMBO | Yes | Target model architecture (FLUX1.D, FLUX2.D, SDXL) |
|
||
| `batch_size` | INT | No | Number of latents in batch (1–64, default: 1) |
|
||
| `clip_multiplier` | COMBO | No | Multiplier for target CLIP resolution (1x, 2x, 4x, default: 1x) |
|
||
|
||
### Outputs
|
||
|
||
| Parameter | Type | Description |
|
||
|-----------------|--------|-----------------------------------------------|
|
||
| `LATENT` | LATENT | Generated latent tensor |
|
||
| `RND_NOISE` | NOISE | Noise for the sampler |
|
||
| `SEED` | INT | Passthrough of latent seed |
|
||
| `NOISE_SEED` | INT | Passthrough of noise seed |
|
||
| `MEGAPIXELS` | STRING | Passthrough of selected megapixels value |
|
||
| `WIDTH` | INT | Calculated latent width in pixels |
|
||
| `HEIGHT` | INT | Calculated latent height in pixels |
|
||
| `TARGET_WIDTH` | INT | CLIP target width (WIDTH × clip_multiplier) |
|
||
| `TARGET_HEIGHT` | INT | CLIP target height (HEIGHT × clip_multiplier) |
|
||
|
||
### Usage
|
||
|
||
Generates a ready-to-use latent and noise pair with reproducible results via independent `seed` and `noise_seed` controls. Dimensions are
|
||
automatically calculated from the selected aspect ratio, megapixels, and model architecture. Connect `LATENT` and `RND_NOISE` directly to a sampler,
|
||
and pass `TARGET_WIDTH` / `TARGET_HEIGHT` to a CLIP text encoder for conditioning.
|
||
|
||
---
|
||
|
||
[View on GitHub](https://github.com/tenser-tensor/ComfyUI-TenserTensor)
|