Writing effective prompts for better color grades
Good prompts are specific about visual intent, not only adjectives. This gives the generator stronger constraints.
Use structure, not adjectives alone
Instead of 'make it cinematic', provide the exact look components. Think in buckets: tone, color balance, contrast shape, and texture.
Structure prompts in short blocks so the model can anchor each style dimension separately.
- Start with one sentence on mood and finish with technical expectations.
- State what should not change first (skin tone, black levels, saturation caps).
- Add intensity and contrast instructions last.
- Iterate by adding one constraint at a time.
- Scene type: street, portrait, interior, food, cinematic filmic.
- Primary mood: warm, cool, muted, vibrant, natural.
- Tone map: highlight retention, black lift, toe point, contrast curve.
- Specific output: DaVinci Resolve `.cube`, Premiere `.3dl`, Lightroom `.xmp`.
What to avoid in early prompts
Overly broad instructions usually produce generic outputs. Multiple competing direction statements can pull grades in opposite ways.
Avoid saying both high contrast and flat at the same time unless you want a very specific look.
A reliable prompt template
Use this baseline and then iterate on one line at a time.
- Mood + lighting style.
- Color temperature and saturation direction.
- Contrast behavior and shadow detail preferences.
- Destination software and file format.
Keep one constraint per sentence and test one variable change per retry.
Common Questions
How many characters should a prompt be?
There is no strict minimum. Short prompts work, but 120–300 characters usually give enough control for first output.
Should I describe skin tone explicitly?
Yes if people are visible. Mentioning 'preserve natural skin tones' usually improves safety for grading.
Related guides
Fix weak, flat, or over-processed LUT outputs
Unexpected output is usually from prompt direction mismatch. Iterative refinement solves most issues without redoing from scratch.
Read guideFrom prompt to LUT: how the workflow works
Your prompt is converted into a structured color grade request, then translated into a LUT profile and Lightroom presets.
Read guidePrompt formulas for scenes, portraits, and products
Different subjects need different grading priorities. A portrait prompt should protect skin tone, while landscape prompts can accept stronger color push.
Read guideReady to generate?
Use these prompts as a starting point and generate in-app for your own imagery.