All skills
hoodini avatar

/image-master

@83e2083

Master prompt-engineer for photoreal, artifact-free AI still images on ANY tool (Reve, Midjourney, Flux, GPT-image, Imagen, Nano Banana, Stable Diffusion). Builds prompts that hit National-Geographic-grade realism — true skin/fur texture (no plastic), correct anatomy/hands/faces, physically coherent light/shadows/reflections, clean legible text — while engineering deliberate visual IMPACT (color contrast, composition, awe/adrenaline, a striking point of view). Runs a gated process: lock the point-of-view and build the 8-block Capture Stack BEFORE generating, then run a forensic pre-submit inspection against the known artifact list. Use whenever the goal is a single still image that must look REAL and hold up under close inspection — contest entries, hero shots, product/character/wildlife/architecture/concept art, or any time AI images come out plastic, distorted, or fake. Composes with nano-banana-2 (execution) and director (motion). Hebrew triggers: תמונה, תמונות, פוטוריאליזם, ריאליזם, לייצר תמונה, פרומפט לתמונה, בלי עיוותים, עור פלסטיק, ידיים מעוותות, פרצוף מעוות, חדות, נשיונל ג'יאוגרפיק.

Use this Skill: https://skilld.dev/gh/hoodini/ai-agents-skills/image-master

This session only. Nothing lands on disk.

references08-series-consistency.md

≈1.8k tokens on demand. Your agent reads this file only when SKILL.md points to it.

08 — Series Consistency: Cohesion + Recurring Characters

When the deliverable is a SET (contest entry, campaign, photo-essay), two problems appear that a single image never has: the set must read as ONE author (cohesion), and any recurring subject must stay the SAME subject (character consistency). Plus the hard rule that governs all output.

0. The full-prompt hard rule (non-negotiable)

Every prompt is 100% complete and copy-paste-ready. NEVER "the previous prompt plus…", "same as above but…", "[insert]", "(keep the rest)", or any abbreviation. If two prompts are 90% identical, write both out in full — repetition is correct; a reference-back is a defect. The user copies one self-contained block per image into the tool; a shortcut breaks the workflow.

1. Series cohesion — make 10 images read as one body of work

Lock a small visual DNA and weave it verbatim into EVERY prompt (the "Style Lock"):

  • ONE film stock / capture signature (e.g. CineStill 800T for a night series — its warm halation around bright lights unifies the set and is on-theme).
  • ONE grain / texture treatment.
  • ONE lighting principle (e.g. a single motivated warm source against a cool night).
  • ONE color logic / grade (teal-orange, deep protected blacks, muted editorial).
  • ONE aspect-ratio family (vary within it for rhythm — 3:2 / 4:5 / 16:9 — not random).

Cohesion lives in the Style Lock; range lives in the treatment (lens, distance, time, energy). That separation is how a single-theme set scores Range without falling apart.

2. Character consistency — keep the SAME animal across images

Ranked by reliability:

  1. Reference image (strongest, tool-native). Generate one "hero reference" of the character first, then feed it into every generation. Reve references are first-class and lossless across edits; Nano Banana takes up to 14 refs; Midjourney --cref; SD IP-Adapter.
  2. Reve layout-node reuse. Reve is layout-first — build the character once and reuse/duplicate that node across compositions, editing pose/scene around it. Surest path within one tool.
  3. Verbatim Character ID block. A fixed, fully-written description with 2–3 UNIQUE identifiers (a notched ear, a specific scar, exact eye color) pasted identically into every prompt. Unique marks stop "generic cheetah" drift. Most portable across tools; aligns with the full-prompt rule.
  4. Seed lock (SD/MJ) for stylistic stability; LoRA / fine-tune (SD) for bulletproof consistency if you have training images.

Best practice: reference image (1) + verbatim ID block (3) together.

3. Worked series — "Through Their Eyes: Israel at War" (10 images)

Unified theme: life in Israel during war told through animals' helplessness — reflecting what people feel. Honest Range bet: a single-theme set risks the Range criterion; we win it by maximizing treatment range (discipline, lens, distance, time, energy) while the Style Lock holds cohesion. Recurring protagonists (HOPE, MARCUS) tie it to Yuval's Midbarium "Born to Run" characters.

The Style Lock (paste into every prompt, verbatim)

shot on CineStill 800T 35mm film, subtle film grain and faint halation glow around bright lights; a single motivated warm light source (Iron Dome airburst / emergency strobe / bare bulb) against a cool desaturated night; warm-cool teal-orange separation, deep protected blacks, muted editorial photojournalism grade; natural fur and skin texture with real micro-detail, one sharp catchlight per eye, no plastic sheen.

Character ID blocks (paste verbatim wherever they appear)

HOPE — an adult female cheetah, lean and athletic, tawny-gold coat with dense round black spots, the signature black tear-stripe lines running from the inner corners of her amber eyes down to her mouth, a small notch in the upper edge of her right ear, a faint old scar on her left foreleg.

MARCUS — a mature male white lion, pale cream-white coat, heavy blond-white mane with faint amber tips, pale gold-green eyes, a thin old scar across the bridge of his broad muzzle.

The ten (each anchored to a REAL, researched Israeli home-front pain — see references/10)

All 4:5 magazine-cover format; masthead "THROUGH THEIR EYES".

# Headline Animal The real pain it illustrates Treatment Risk
1 THE WITNESSES Bengal tigers (one turns) helpless watching the interception decide your fate telephoto night silhouette, war-in-eye low
2 THE WAKING lioness over sleeping cubs the dilemma: wake the just-asleep kids for the siren, or gamble? intimate den interior cub faces (closed eyes)
3 DON'T WAIT FOR ME old greying lion the elderly/disabled who can't reach shelter in time environmental, aged subject aged face large
4 A TIN OF SARDINES mixed animals crammed the public shelter — no privacy, no toilet, strangers at 3am cramped group interior multiple faces
5 NOWHERE TO GO doe + two fawns at a bare wall 56% of homes have no safe room exposed-family medium small faces
6 TOO LATE HOPE + cub the 15–90 second window to reach shelter panning action blur paws, open mouth
7 THE SHIELD MARCUS over cubs, struck shelter hit despite doing everything right; the body as last shelter dust/aftermath protective cubs, dust restraint
8 THE WATCH exhausted mother, cubs asleep bone-deep exhaustion; the parent who performs calm, never sleeps low-light intimate weary face large
9 THE SOUND a hare frozen mid-flinch the siren as trauma trigger; hypervigilance, phantom sirens prey-freeze, ears straining face/ears
10 DO YOU SEE ME? street dog, direct gaze the universal plea; pets paralyzed by terror extreme close portrait symmetry

Emotional arc: helpless watching → the waking → the left-behind → the crush → the exposed → the race → the failed shelter → the sleepless watch → the trigger → the plea. Treatment still spreads across silhouette / interior / environmental / group / action / portrait for the Range criterion. Benched alternates (all pain-valid, swap in freely): baboon "caught in the open, cover your head"; lone oryx "one life beneath a war"; sunbird "her wings the only roof"; gazelle dawn "the morning after"; stork flock "exodus / displacement".

Full standalone prompts are produced from this bible using the Style Lock + Character ID blocks + the 8-block stack; each must pass references/07 at 100% zoom before it joins the set.

4. Submit-set checklist

  • All 10 carry the Style Lock verbatim (cohesion holds).
  • Recurring characters use the same reference image + the verbatim ID block.
  • Treatment ranges across discipline/lens/distance/time/energy (Range hedge).
  • Each image individually passes the references/07 forensic pass at 100%.
  • No two images are near-duplicates; each earns its slot.
  • Every prompt handed over is complete — zero shortcuts.

Source: SKILL.md on GitHub

No alerts3mo3 checks · Risk SAFE
  • Gen Agent Trust Hub3mo

    The skill is a comprehensive instructional framework for AI image prompt engineering. It consists entirely of Markdown documentation and does not contain any executable code, scripts, or mechanisms for data exfiltration. All external references are informative and point to legitimate AI and photography resources.

  • Socket3mo

    No alerts

  • Snyk3mo

    Risk: LOW · No issues

Signed by skilld at 83e2083. This ties the file your Agent reads to that commit on GitHub. It does not review the instructions.

Last checked against GitHub 2 days ago.

Steadyupdated 4 months ago

README badge

README badge for hoodini/ai-agents-skills/image-master