ComfyUI Node
Latent → CV Array
Unwraps a ComfyUI LATENT ({'samples': [B,C,H,W]}) back to a float ndarray [H,W,C] so it can be processed by OpenCV nodes. Channels move to the last axis. Values are kept as float32 - never quantized. This is the inverse of 'CV Array → Latent'.
Latent → CV Array
- latents
- nparray
◄batch_index0►
Categoryimage/CV/low-level
Inputs (2)
| Name | Type | Default | Description |
|---|---|---|---|
| latents | LATENT | A LATENT dict with 'samples' tensor of shape [B,C,H,W]. The selected batch index is extracted and channels are moved to the last axis, yielding [H,W,C]. | |
| batch_index | INT | 00–4095 | Which latent of the batch to convert. Clamped to the batch size. |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| nparray | NPARRAY | Float32 ndarray [H,W,C] (C = latent channels; 16 for qwen_image_vae, 4 for SD 1.5 / SDXL). |