Icy RepeatLatentBatch

IcyRepeatLatentBatch takes each latent in a batch and repeats it a set number of times. Sounds trivial; it's actually the plumbing that makes a bunch of real workflows work. It's how you stretch one image into eight identical copies so a downstream pass can treat them all separately - different prompts, different denoise, different seeds - and it's how you pad batch sizes for video and batch-processing pipelines.

The "Icy" prefix is the pack's privacy coat, not a different repeat. IcyHider subclasses core ComfyUI nodes at startup - same inputs, outputs, and behavior, filed under the IcyHider Comfy Core category - and its JavaScript hides previews until hover. This is core RepeatLatentBatch with a cover on. If you know the core node, you know this one.

The two inputs

From the schema:

• samples - the LATENT batch to repeat.
• amount - how many times each item is repeated, default 1, up to 64. amount: 2 on a batch of 4 gives you 8 latents: [a, a, b, b, c, c, d, d].

One output: LATENT, the expanded batch.

Where it actually matters

The textbook case: you've generated one image and want to try a second pass over it with several denoise strengths or prompts. Rather than re-roll the base image for each variation, repeat the single latent to batch size N, then branch it into N sampler setups with different settings. Everything downstream sees the same starting latent, so your comparisons are apples-to-apples instead of seed-lottery.

It's also the piece that makes batch semantics explicit. ComfyUI batches flow left to right, and a lot of nodes assume matching batch shapes - if node A produces 4 latents and node B expects 8, RepeatLatentBatch is the adapter. Same trick applies in the other direction with its sibling, LatentFromBatch, which carves a batch back down.

One thing to keep straight: repeating isn't generating. You get N identical copies, not N variations - you have to actually run them through different conditions to get different results. A common beginner confusion is expecting amount: 4 to give four different images.

Install and gotchas

Frontend-only pack, no dependencies:

# ComfyUI Manager: search "IcyHider", Install, restart

# or:
cd ComfyUI/custom_nodes
git clone https://github.com/icekiub-ai/ComfyUI-IcyHider.git
# restart ComfyUI

Pack-wide: the cover is DOM-based, so the newer Nodes 2.0 canvas frontend may not hide previews, and the hiding is screen-only - nothing on disk changes. And if your repeated batch is feeding a sampler, remember each copy costs sampler time and VRAM; a batch of 16 at high res gets expensive fast.