AutoAdjust Suite

AutoAdjustNode gives photos a quick “auto levels / auto tone / auto color” pass similar to Photoshop. AutoColorConfigNode is the helper dial that lets you toggle those same options across several branches at once.

AutoAdjustNode

Inputs

image – The picture you want to clean up; feed it from any loader or crop.
precision – Choose Histogram (fast) for everyday batches or Exact when you notice uneven results and want a pixel-perfect read.
auto_levels – Stretch shadows and highlights so the full brightness range is used.
levels_shadow_clip_pct – Percentage of darkest pixels to shave off while stretching. Lower = gentler.
levels_highlight_clip_pct – Same idea for the brightest pixels.
levels_gamma_normalize – Gently nudge midtones toward middle grey so portraits don’t look flat.
auto_tone – Adds a second balancing pass that adjusts each colour channel separately.
tone_mode – Per-channel boosts each colour independently for punchier contrast; Monochromatic keeps colour relationships intact.
tone_shadow_clip_pct / tone_highlight_clip_pct – How aggressive that second pass should be in shadows and highlights.
auto_color – Remove colour casts (too cool, too warm) automatically.
snap_neutral_midtones – Focus the colour fix on midtones so highlights don’t overpower the adjustment.
flip_horizontal – Mirror the image left/right without adding another node.

Output

image – The adjusted photo, clamped to [0, 1] and ready for your next node.

AutoColorConfigNode

Use this helper when you want multiple AutoAdjustNode blocks to respond to the same switches.

Inputs / Outputs

auto_levels
auto_tone
auto_color
flip_horizontal

Each input is a simple checkbox, and each output mirrors that value. Wire the outputs straight into matching inputs on any AutoAdjustNode instances you want to keep in lockstep.

Where It Fits

Drop AutoAdjust at the top of a portrait pipeline when you need a neutral base grade before creative work. It also shines when batch-cleaning camera JPGs with mixed lighting or prepping reference frames for diffusion and upscaling jobs.

Tuning Tips

Lower the clip percentages when you must protect detail in highlights or shadows.
Switch to Exact precision for small batches, archival scans, or whenever histogram estimates wobble.
Enable snap_neutral_midtones for tungsten or mixed indoor lighting to keep skin tones believable.
Use the flip toggle during QA passes to spot composition issues or prep mirrored training data.

Troubleshooting

Image looks flat – Try disabling auto_tone or reduce the tone clip percentages.
Colours drift too far – Turn off auto_color, or enable snap_neutral_midtones for a gentler fix.
No visible change – Confirm auto_levels is on and both clip percentages aren’t set to zero.

See docs/screenshots/auto_adjust_node.png for the widget layout inside ComfyUI.

3.0 KiB Raw Blame History Unescape Escape