AI Illustration Artist

Basically: I need someone to take photos of buildings and turn them into image to image colorized cartoonish versions. The ideal candidate should have experience in architectural photography and image processing, with a keen eye for detail and creativity to achieve a cartoonish effect. The project involves transforming real-world images into vibrant, stylized versions that capture the essence of the buildings. ───────────────────────────── THE WORK IN ONE SENTENCE ───────────────────────────── Using ChatGPT (gpt-image-1) or Claude image tools, design a pair of mascot characters for each project and produce one flat-color, bold-outlined, children's-illustration scene: a REAL building supplied by the client, its surroundings, and the characters — delivered as clean PNGs that drop into our layout template without retouching. You are producing the artwork that sits inside an existing frame. The layout, typography and everything around the picture are already built and are not yours to change. ───────────────────────────── WHAT YOU GET FOR EACH PROJECT ───────────────────────────── - The property name, address, and reference photos of the real building - Brand colors and any character direction (animal, outfit, local tie-in) ───────────────────────────── PART 1 — CHARACTER DESIGN ───────────────────────────── Each project gets a pair of original characters — one larger, one smaller — in the same flat, bold-outlined children's-illustration style as the scene. They are usually animals, dressed in something that ties to the property or its location. - Design 3-4 candidate pairs per project for selection, then refine ONE - The pair must read as two distinct characters, not twins at two sizes: different size, different color, different outfit detail - Made for young children: big head, simple shapes, friendly face, no scary features, nothing sharp - Clean enclosed shapes, bold uniform black outline, no shading inside - Generate on a transparent background so the pair can be composited onto any scene - Once a pair is APPROVED it becomes a fixed asset. From that point it is reused by compositing the approved artwork — never re-generated, never restyled, never re-posed without asking. Character drift between images is the single most visible failure we have. Deliverable per project: - characters1.png both, transparent background - characters2.png each one alone, transparent - A one-line character note: names, species, outfit, tie-in ───────────────────────────── PART 2 — THE SCENE ───────────────────────────── Deliverable per project: 1. scene-project.png 1536 x 1024, RGB, no alpha, no text overlays — building, surroundings and characters composed 2. plate-project.png the same scene with NO characters (empty foreground where the characters stand) Files named exactly as specified, delivered in one folder, PNG only. Two revision rounds are included per project. ───────────────────────────── STYLE RULES (non-negotiable) ───────────────────────────── These are the rules that get work rejected. They are absolute. THE REAL BUILDING - The client's actual building at its actual address. Never substitute a different landmark, never "improve" the architecture, never move it to a different setting. Work image-to-image from the supplied photos (gpt-image-1 `images.edit`, input_fidelity high) — text-to-image invents a different building every run. FLAT COLOR, BOLD OUTLINE - Every shape carries a uniform bold black outline. Flat fills only. - NO drop shadows, NO gradients, NO painterly shading, NO haze. - Colors sampled from the real reference photos, not eyeballed. SKY - Exactly 4 clouds unless the brief says otherwise. All clouds treated identically — either all outlined or all un-outlined, per the brief. - Cloud interiors are one flat cream. No color banding inside a cloud. CHARACTERS - Build the scene as a character-free plate first, then composite the approved characters onto it. Do not ask the model for building plus characters in one image — it will not respect character size or placement. - Zoom in on every character at 100% before sending. AI art routinely adds or drops small details between generations; you are responsible for catching it, not us. BLANKS, NOT ERASURES - If the brief calls for a sign panel, flag or blank plate, GENERATE it as an empty panel in the artwork. Do not paint something and erase it later — every erasure leaves a smear and we can see it. CLIENT MARKS - Never draw, letter or approximate a client's logo. Leave the blank; we place the licensed artwork ourselves. TEXT - No text anywhere in the scene unless the brief explicitly asks for it. ANYTHING COUNTABLE - Windows, stripes, floors, columns: if the brief gives a number, the number is right or the piece comes back. ───────────────────────────── YOUR SELF-QA BEFORE SUBMITTING ───────────────────────────── Run this every time and state that you ran it: [ ] Building matches the reference photos — shape, floors, roofline, color [ ] Cloud count and cloud outline treatment consistent across the sky [ ] No shadows, no gradients, no washed-out haze [ ] Every character inspected at 100% zoom, details correct and consistent with the approved character artwork [ ] Every required blank panel is present and truly empty [ ] No logo lettering, no stray text [ ] Nothing important within 6% of any edge (the frame crops the border) [ ] Exact pixel size, correct filenames, PNG ───────────────────────────── WHO THIS IS FOR ───────────────────────────── - You are fast and fluent with gpt-image-1 / ChatGPT image generation, especially image-to-image editing from a reference photograph - You have an eye for character design that appeals to young children - You can read a reference image and match a house style exactly, rather than expressing your own - You are comfortable with light compositing (Photoshop, Affinity, or Python/PIL) to paste a transparent cutout onto a plate at a specified position and scale - You check your own work obsessively and would rather send one correct file than three fast ones - Bonus: basic Python, so you can run our image-cleanup scripts yourself before delivery Not required: traditional illustration skill, drawing ability, a degree. This is a prompting-and-QA craft job, not a fine-art job. ───────────────────────────── HOW WE WORK ───────────────────────────── - Batches of 2-6 projects at a time, ongoing - Turnaround: 48 hours per project from brief to first delivery - Two revision rounds included; further rounds only if the brief changed - All output is work-for-hire, assigned to CrayonStories LLC - You may not reuse, publish, or portfolio the characters or the artwork without written permission ───────────────────────────── HOW TO APPLY ───────────────────────────── Send: 1. Two examples of AI image work where you had to match an existing style or reproduce a real place accurately (reference/result pairs) 2. One example of a character you designed, shown in at least two different images, so we can see how well the identity held 3. One paragraph: how you would keep a real building's facade from drifting across generations 4. Which image tools you use, and how you edit rather than regenerate