What is Tiled Diffusion?

Tiled Diffusion (MultiDiffusion / Mixture of Diffusers) splits a large canvas into overlapping tiles, denoises each tile separately, and fuses them. This lets you generate images far larger than your VRAM would normally allow because only one tile's worth of latent is processed at a time.

How do I choose between MultiDiffusion and Mixture of Diffusers?

MultiDiffusion is faster and works well for moderate upscales. Mixture of Diffusers blends tile boundaries more smoothly and is the better choice for very large canvases where visible seams are a risk. This tool recommends one based on your target megapixels.

Why do I still run out of memory?

Tiled Diffusion handles the UNet, but the VAE decode at the end can still spike memory on a full-size latent. Enable Tiled VAE alongside Tiled Diffusion so the decode is also tiled, and keep batch tile count at 1.

What overlap should I use?

48-64px is a good default. Too little overlap produces visible seams between tiles; too much wastes compute by re-processing pixels. Increase overlap if you see grid-line artifacts in the output.

Tiled Diffusion Settings Guide

Generating high-resolution images with Tiled Diffusion

Your GPU’s VRAM caps the resolution you can generate in a single pass — but Tiled Diffusion removes that ceiling by splitting the canvas into overlapping tiles, denoising each one independently, and fusing the results. This tool calculates a tile size that fits your VRAM, works out how many tiles your target resolution needs, and tells you whether MultiDiffusion or Mixture of Diffusers is the better method for that canvas.

How it works

You enter the target resolution, your available VRAM, and a preferred tile overlap. The tool maps your VRAM to a safe latent tile size (smaller tiles for 4-6GB cards, up to ~1920px for 24GB). It then computes the tile grid from your target dimensions and the effective stride (tile size minus overlap), giving you the total number of tiles per image and an estimate of relative cost. Larger canvases trigger a Mixture-of-Diffusers recommendation for cleaner seams.

Tips for clean, fast tiling

Pair with Tiled VAE. The biggest memory spike is often the final VAE decode — tiling it too prevents an out-of-memory crash at the last step.
Keep batch tile count at 1 on low VRAM. Higher batches process tiles in parallel but multiply peak memory.
Raise overlap if you see seams. Grid lines or repeated edges mean tiles aren’t blending — bump overlap from 48 to 64-96px.
Use it for upscaling, not just generation. Tiled Diffusion combined with a ControlNet tile model is a powerful high-res-fix and upscaling workflow.