Latent Noir

Independent image lab

AI image experiments, with the settings attached.

Local ComfyUI runs with FLUX.3 Klein and Moss Image. Every experiment includes its prompt, sampler, CFG, seed and 4K crops.

Experiments
14
Models
02
Categories
08
Files
42
Updated 2026-10-2597% smaller than source rendersOne prompt, both models

Two models, different jobs

One prompt, both models, same seed and canvas.

Compare

Primary image model

FLUX.3 Klein int8

7 experiments

Distilled FLUX.3 image model, int8. The sharper pass for neon, chrome, hard light, space and macro detail.

Sampler

euler / simple

CFG

1

Steps

16

Atmosphere model

Moss Image int8

7 experiments

Moss Image SFT, int8. Darker, softer rendering for weather, fog and negative space.

Sampler

er_sde / beta

CFG

4

Steps

50

How a run is published

01

Prompt

A specific subject, composition, light and material brief — plus a separate negative prompt.

02

Render

Local ComfyUI run. The model, sampler, scheduler, steps, CFG, guidance and seed are recorded.

03

Publish

One 16:9 master, attention-cropped for phone and ultrawide, then encoded to AVIF and WebP.

Browse the full archive

42 crops, every prompt and every seed.

All experiments