Tile Test: Render

A ComfyUI node in image/upscaling/tile testing with 19 inputs and 2 outputs.

By Blakeem·Created 2 months ago·Updated 3 days ago· 25

Tile Test: Render

image
layout
model
clip
vae
captions
negative

tiles
blocks

◄seed0►

◄sampler_namedpmpp_2m►

◄schedulersgm_uniform►

◄steps20►

◄cfg3.5►

◄denoise0.50►

◄anchor_sourcesource image►

◄surfacevision tokens and captions►

◄canvas_tokens165►

◄crop_tokens165►

◄tiles►

◄with_neighborstrue►

Categoryimage/upscaling/tile testing

Inputs (19)

Name	Type	Default	Description
image	IMAGE		The canvas from Tile Test: Upscale, at the layout's target size.
layout	CATR_LAYOUT		The layout from Tile Test: Layout. Every rect this node samples is read from it.
model	MODEL		The diffusion model that denoises each tile.
clip	CLIP		Must be a vision-language text encoder (Krea 2 family). There is no positive prompt input, since each tile is conditioned on the image itself.
vae	VAE		The VAE that encodes and decodes each tile.
seed	INT	00–18446744073709550000	Noise is drawn once for the entire image and then sliced for each tile.
sampler_name	COMBO	dpmpp_2m	The synchronized tile engine supports these samplers: euler, dpmpp_2m, heun, dpm_2, exp_heun_2_x0, dpmpp_2m_sde, dpmpp_2m_sde_gpu, dpmpp_2m_sde_heun, dpmpp_2m_sde_heun_gpu, exp_heun_2_x0_sde. Any other sampler is rejected.
scheduler	COMBO	sgm_uniform	How the noise levels are spaced across the steps.
steps	INT	201–10000	The number of sampling steps each tile runs at any denoise value.
cfg	FLOAT	3.50–100	How strongly each tile follows its positive conditioning over the negative.
denoise	FLOAT	0.500–1	The fraction of the noise schedule each tile runs, where 0 samples nothing and returns the image unrefined.
anchor_source	COMBO	source image	What fills context_anchor. source image keeps the result true to the input. live canvas adds more detail and drifts further from the input.
surface	COMBO	vision tokens and captions	What fills every tile's positive. The two caption surfaces need Tile Test: Captions connected. The VL nodes use the vision tokens and captions surface by default.
canvas_tokens	INT	1650–1953	Vision rows a tile takes from one encode of the entire image. 0 turns that source off.
crop_tokens	INT	1650–1953	Vision rows a tile takes from the encode of its own crop. 0 turns that source off.
tiles	STRING		Comma separated tile numbers, as Tile Test: Layout labels them. Empty renders the entire canvas as one run.
with_neighbors	BOOLEAN	true	On renders the tile with its bordering tiles as lanes of one run, which is what the full run does. Off renders the tile alone, and its anchor ring then stays unrefined source, since no neighbour lane refines it.
captionsopt	CATR_CAPTIONS		The captions from Tile Test: Captions. The two caption surfaces need them and the vision tokens surface ignores them.
negativeopt	CONDITIONING		Optional negative conditioning. Unconnected it is an empty encode of this node's CLIP.

Outputs (2)

Name	Type	Description
tiles	IMAGE	One image per named tile, the refined canvas cut at the tile's crop rect. An empty tiles input returns the entire refined canvas.
blocks	IMAGE	One image per named tile, the refined block rendered for it, which spans its bordering tiles when with_neighbors is on. An empty tiles input returns the entire refined canvas.