Multiply Latents
Elementwise product of two latent tensors.
Category: ModularDiffusion/Transform
TL;DR
output = left_latent * right_latentelementwise.- For masked compositing of two different latents, prefer Latents Composite Mask — it's purpose-built and handles the mask resampling for you.
Typical workflow position
Latent A ─┐
├─→ [Multiply Latents] → Add Latents / Generate
Latent B ─┘
Node preview

Inputs
| Name | Type | Required | Notes |
|---|---|---|---|
left_latent |
LatentArtifact |
Yes | |
right_latent |
LatentArtifact |
Yes | Must match left_latent shape. |
Outputs
| Name | Type | Notes |
|---|---|---|
output_latent |
LatentArtifact |
left * right. |
Tips & pitfalls
- Elementwise, not matrix multiplication. Each element in
ais multiplied by the corresponding element inb. - Most useful for scaling. Multiplying one latent by a near-constant tensor is the most predictable use; combining two arbitrary latents elementwise tends to produce unexpected results.