ZaddyTools

Character Sheets

Drop these into Higgsfield, Sora, Veo, Runway, or Midjourney as character refs.

8

Click any topic to copy its prompt. Paste into Midjourney, Sora, Veo, Runway, Flux, or Nano Banana.

5

Match the ratio to your platform before you generate. Click any to copy.

7

Battle-tested prompts that produced standout results. Click any to copy.

0

No saved examples yet.

Skeleton · platform quirks · pre-send checklist. The shortest path from idea to paste-ready prompt.

Image

[SHOT TYPE] of [SPECIFIC SUBJECT] [doing ACTION], in [ENVIRONMENT], [LIGHTING], [STYLE], --ar [RATIO]

Video

[CAMERA MOVE] of [SPECIFIC SUBJECT] [verb-ing], in [ENVIRONMENT]. [LIGHTING]. [MOOD]. [LENS / STYLE]. Duration: [Xs].

Subject is specific — "weathered 60-year-old fisherman" beats "a man".
Lighting is stated — time of day or quality of light.
Aspect ratio matches the platform — 1:1 / 9:16 / 16:9.
Reference image is attached if character consistency matters — text alone won't lock a face.
Right model for the job — generate vs edit vs video. Instruction vs descriptive grammar.

Multi-reference stacking — combine --cref (character) + --sref (style) + --oref (omni) with weights for a signature look.
Flux LoRAs — train once on 15–20 images, deploy across every project. Better character lock than --cref drift.
Parameter dials — --stylize 50 (literal) vs 750 (painterly), --weird 1000, --chaos 30 for variation hunting.
Image-to-image evolutionary loop — generate 4, pick best, regenerate from it with prompt tweaks.
Start + end frame interpolation (Runway, Kling, Luma) — drop two stills, model fills the motion.
Skip negatives on modern models — Flux, Sora 2, Veo 3 handle "no extra fingers" intrinsically. Over-specification hurts.