Color grading, without the jargon wall.

Good grading is controlled decision-making. Correct the image first, shape the look second, then use scopes and comparison tools to make sure the grade still holds together.

1. Correct exposure before adding a look

Exposure moves the whole image. Highlights, shadows, whites and blacks shape different tonal regions. Use the RGB histogram and waveform to catch clipping, but judge the actual image before chasing a technically perfect graph.

2. Neutralize unwanted color casts

Temperature moves the image between cooler and warmer. Tint moves between green and magenta. Neutral surfaces, clouds, white clothing and believable skin are useful references. Small corrections usually beat dramatic ones.

3. Shape contrast with curves

The master curve changes overall tonality. Red, green and blue curves can also create channel-specific color shifts. Keep endpoints controlled and use a few intentional points instead of building a jagged curve.

4. Use the color mixer for selective changes

The HSL mixer targets eight color families. Hue changes where a color sits, saturation changes its intensity, and luminance changes how bright that color appears. It is useful for controlling vegetation, skies, clothing and warm skin-adjacent colors without pushing the whole image.

5. Use color wheels for mood

Three-way wheels target shadows, midtones and highlights. A subtle cool push in shadows and a gentle warm push in highlights can create separation without turning the image into an obvious filter.

6. Keep skin protection on when grading people

MintGrade uses a conservative pixel-level skin likelihood mask to reduce aggressive creative color shifts in likely skin regions. It is a guardrail, not face recognition and not a perfect semantic mask, so always judge the actual result.

7. Read Grade Health as a warning system

Grade Health checks for clipped highlights, crushed shadows, unusually extreme color and likely skin-tone problems. A high score does not mean the image is artistically correct; it means fewer common technical warning signs were detected.

8. Match references intelligently

Reference Match compares broad luminance, contrast, saturation and channel balance between your source and a chosen reference. It produces editable controls rather than baking in a black-box filter. It is intentionally conservative.

9. Compare and save versions

Use the split comparison to keep perspective on the original. Save manual versions when you reach a meaningful direction instead of relying only on undo history. Normal project edits autosave locally, so Versions are for creative checkpoints rather than basic persistence.

Why local processing matters

The current grading workflow runs in your browser with WebGL2. Images are not intentionally uploaded to an application server for editing. Saved projects use IndexedDB on the current device and browser profile.

10. Export from the original

The interactive preview is resized for speed. Export re-renders from the original decoded image with overlapping GPU tiles when needed, then optionally downsizes to a Web, Social or Quick-share recipe while keeping the same grade and aspect ratio.

A reliable order of operations

1. Load the image.
2. Correct exposure and tonal extremes.
3. Fix temperature and tint.
4. Shape contrast.
5. Apply an adaptive look or reference match if useful.
6. Refine HSL and color wheels.
7. Use curves only where they add control.
8. Check Grade Health and scopes.
9. Compare against the original.
10. Save a creative version when useful; the project itself autosaves.
11. Export at original resolution.