Krea 2 (bf16) in ComfyUI: 2048×2048 with chunked execution

A text-to-image workflow for Krea 2 that renders at 2048×2048 in 8 steps. It uses the LayerStream: Krea2 Chunking node and a tiled VAE decode so large images don't hit the memory wall.

What problem it solves

Big image models have two memory costs: the weights, and the activations. The activations grow with image size. For a full-size model like the 24.5 GB bf16 Krea 2, ComfyUI streams the weights, but the activations at large sizes can still spill into slow shared memory.

The node runs the transformer in slices, so the activation peak stays small. Decoding with tiles does the same for the VAE step.

Real-world result

On a 24 GB card with default launch flags, one production run of the bf16 model took about 30 seconds with the node. Without the node, it did not finish in a reasonable time. That is one run on one machine, so treat it as a report, not a guarantee.

Does it change the picture? Please read this

Not bit-identical. I won't claim that for Krea 2, because it isn't true:

• With a single chunk the output was bit-identical.

• With several chunks, the bf16 math rounds slightly differently depending on how many rows each slice has. The output is numerically equivalent but not pixel-identical. I measured about 40.7 dB PSNR against the stock render at 1 step, about 29–30 dB at 8 steps (1024²) and 38.5 dB at 4 MP. Compare a few renders side by side yourself, since I can't judge your taste for you.

• I did not test whether the int8 or fp8 versions are exact. The tests above are for bf16.

Speed and limits

• Slower: chunking costs time because weights are re-fetched for each slice. In my hard-cap tests it was about 2.7 to 4 times slower than stock where the stock run fit at all.

• At 1 MP, the launch flags alone were enough. The node isn't needed there.

• At 4 MP the node lowered PyTorch's memory floor by about 1 GB (stock worked at 3 GB, the node at 2 GB, both with the tiled VAE). That is PyTorch's own allocations, not total GPU usage, and not a promise for small cards.

• Tested only on Windows with an RTX 4090 (24 GB), ComfyUI 0.39.0.

• The node patches ComfyUI's Krea 2 code, so a future ComfyUI update could break it.

Requirements

• ComfyUI 0.39.0 or newer

• The osiworx-pack custom node pack, from ComfyUI-Manager or https://github.com/osi1880vr/osiworx-pack

• Krea 2 model, the Qwen3-VL 4B text encoder (fp8) and the Qwen Image VAE. The file names in the loaders are from my machine, so pick your own. The workflow's default name is krea2_turbo_int8_convrot.safetensors, but my measurements were on the bf16 file.

• Model files are not included.

Measurements and untested areas are in the README: https://github.com/osi1880vr/osiworx-pack