The workarounds and prompt techniques for legible words, logos, or signs in generated images — the hard part.
midjourneychatgptgeminiadvanced
The prompt
Help me get legible text into an AI-generated image.
What the image is & the exact text I need in it: {{describe + the words}}
Tool: {{Midjourney / DALL-E / etc.}}
Give me: 1) THE REALISTIC EXPECTATION — where each tool is decent vs. hopeless with text, and how much text is safe to attempt. 2) THE PROMPT TECHNIQUE — how to phrase it (quoting the text, keeping it short, specifying style/placement) to maximize legibility. 3) THE FALLBACK — when to generate the image WITHOUT text and add it after in a design tool instead (often the pro move). 4) THE PROMPT for my case, plus a text-free version to overlay. 5) The tweaks if the text comes out garbled. Legible over 'AI tried its best'.
578 copies
Why this works
Text is where image models still routinely fail; knowing when to attempt it in-prompt versus generating text-free and adding words after saves hours of fighting garbled letters for a result a design tool does cleanly.
Shared by Sam Kowalski — Full-stack dev. Debugs out loud so you don't have to.
Act as a commercial photography director planning a hero shot.
My product: {{what it is}}
Brand mood: {{3 adjectives — e.g. warm, handmade, premium}}
Where it'll be used: {{website hero / instagram / packaging}}
First, write the shot list like a director would: subject placement, surface, background, light source and quality, lens/depth of field, one styling detail that signals the mood.
Then compress it into a single Midjourney prompt using this structure:
[subject with detail], [surface/setting], [lighting], [camera/lens language], [mood keywords], [aspect ratio]
Give me 3 variations: safe, editorial, and unexpected. Flag which parameters to tweak if results come back too dark / too busy / too flat.
Help me write an image prompt for a product mockup.
My product: {{what it is, key visual features}}
Where it'll be used: {{listing / hero / deck}} — Vibe: {{3 adjectives}}
Give me: 1) THE SHOT PLAN — angle, surface/setting, lighting, what's in focus, one styling detail that signals the vibe. 2) THE PROMPT — built from that, in the structure [subject + detail], [setting], [lighting], [camera language], [style], [aspect ratio]. 3) THREE VARIATIONS — clean/minimal, lifestyle/in-use, and bold/editorial. 4) THE FIX-ITS — what to change if the result looks too flat, too busy, or fake. Optimize for 'I'd click buy', not 'pretty'.
My image prompt isn't working. Diagnose and fix it.
My prompt:
{{paste it}}
What I got vs. what I wanted: {{describe the gap}}
Tool: {{Midjourney / DALL-E / etc.}}
Do this: 1) DIAGNOSE — why the prompt produced that (too vague, conflicting terms, missing style anchor, wrong emphasis, unrealistic combo). 2) THE REWRITE — a cleaner version with the fix, keeping what worked. 3) WHAT EACH CHANGE DOES — so I learn, not just copy. 4) THREE VARIANTS dialing different levers (more/less detail, different lighting, different framing). 5) THE ONE PARAMETER most likely to close the gap. Teach me to fish.