Lesson Content

Diagnosing before changing.

Effective iteration starts with diagnosis: look at the generation result and identify specifically what is wrong before changing anything. Wrong style, wrong lighting, wrong subject rendering, wrong environment, unwanted elements, and quality artifacts each require different fixes. Changing a random prompt element without diagnosis wastes tokens on changes that don't address the actual issue.

Matching the fix to the diagnosis.

Diagnosis-fix mapping: wrong art style -> change style descriptor; wrong mood/atmosphere -> change lighting and mood language; subject doesn't match -> add specificity to the subject description; unwanted element -> add to negative prompt; quality artifact -> add relevant negative descriptor or quality modifier; wrong composition -> add framing and camera direction.

Variations vs. full prompt revision.

Leonardo AI includes variation generation options – producing new images that are variations on the same generation. Use variations when: the generation direction is essentially right but the specific result needs refinement. Use full prompt revision when: the generation direction is fundamentally wrong and requires new descriptors to change it. Variations explore the same direction; revisions change the direction.

Editing vs. regenerating.

When a generated image is close to ideal but has a single specific issue (an unwanted element in one area, a slightly wrong color tone), consider using AI Canvas to edit that specific area rather than regenerating the full image. Targeted editing through AI Canvas preserves what's working while addressing the specific issue – more efficient than regenerating when the core image is strong.