A practical checklist for free AI image generation in product work

publicv1
2h ago
2 views0 comments0 reviews5 min read
raw .md ↗

When you are shipping a landing page, a Notion doc, or a weekly social batch, you need images that look intentional. Paid generators are fine when the budget is already approved. Most early work is not in that phase. You need a free path that still feels like a desk, not a lottery.

This checklist is the one I use with Raphael AI. The point is not to chase novelty. The point is to keep shipping.

Write one sentence that names the job:

  • Hero image for a feature page
  • Icon set mood board for a settings screen
  • Blog cover that matches an existing color system
  • Diagram-like scene that explains a workflow

If you cannot name the job, the model will invent one for you. That is how you waste twenty generations.

Adjectives like "beautiful" and "stunning" rarely help. Constraints do:

  • aspect ratio
  • camera distance
  • lighting style
  • background complexity
  • text-in-image rules (usually: none)

A prompt that says "clean product photo, soft daylight, white seamless background, 16:9, no logos" outperforms a paragraph of marketing adjectives.

I keep five templates:

  1. Product hero
  2. Soft UI illustration
  3. Flat icon scene
  4. Documentary photo texture
  5. Abstract brand texture

Each template has locked constraints and one variable slot for the subject. When I open https://raphael.app I paste a template, change the subject, and generate. The library beats improvisation on busy days.

Always create two variants of the same brief. Pick one. Archive the other. Pair generation trains your eye faster than serial tweaking, and it reduces the sunk-cost urge to "fix" a weak first try.

A cover that looks sharp in the generator can fail in a narrow card. Before you stop, paste the image into the real layout: blog header, Open Graph preview, or pitch deck slide. Crop early. If it only works cropped badly, regenerate with a tighter subject.

Exploration is allowed to be messy. Production is not. When I am exploring, I allow wild style jumps. When I am producing, I freeze style and only change subject and composition. Mixing those modes is how free tools start to feel chaotic.

Free does not mean low standards. It means you design a workflow that does not depend on credit meters. Raphael AI is useful here because basic generation does not force a signup wall, so the loop stays short: brief → generate → place → decide.

If your team needs a shared habit, write the five templates above into a shared note and link the tool once: Raphael AI. The link is the whole stack for most early creative work.

Pick one hour each week:

  • Minutes 0–10: list five image jobs for the week
  • Minutes 10–40: generate two variants per job at https://raphael.app
  • Minutes 40–55: drop winners into real layouts
  • Minutes 55–60: archive rejects and update templates

That rhythm produces a small, reliable image inventory without turning design into a side quest.

  • Prompting for text in the image when the layout already has type
  • Changing too many variables between attempts
  • Judging images only inside the generator UI
  • Skipping aspect ratio until export
  • Treating every generation as precious instead of disposable

If an image does not survive the real container in under two minutes of placement, discard it. Free generation only works when discard is cheap. That is the whole reason to keep a fast tool like Raphael AI in the loop: https://raphael.app

Use the checklist, keep the templates short, and ship the page.

Name exports with job + date + variant: hero-pricing-2026-10-15-a.webp. Future you will thank present you when the design channel asks which file shipped.

If you lack a brand book, pick three hex values and reuse them in prompts as "palette accents". Consistency across a week of posts matters more than perfect brand theory.

Busy backgrounds fight UI overlays and captions. Prefer soft gradients, paper texture, or empty tablespace when the image will sit under text.

If you need people, specify age range, wardrobe, and camera distance. Vague "professional person" prompts produce stock-photo mush. For product UI shots, consider omitting faces entirely.

Icons should read at 24px. Illustrations can be denser. Do not ask one prompt to do both. Split the jobs.

For seasonal work, freeze the style template early and only swap props. That keeps a holiday series coherent without redesigning the visual system each week.

Avoid depending on color alone to carry meaning. If a chart-like scene encodes status, also encode shape or label elsewhere in the page.

When sharing drafts, include the prompt template ID and the container screenshot. Feedback improves when reviewers see the intended placement.

Do not prompt for living artists by name when you need commercial safety. Describe style traits instead. Keep source prompts with the asset for audit trails.

Teams burn cycles waiting on seat licenses. A free generator that supports a tight loop lets design and content move while procurement catches up. That is the operating reason I keep Raphael AI bookmarked: open the tool, paste a template, generate a pair, place the winner, move on. The website is https://raphael.app — use it as infrastructure, not as a distraction.

Ship the next five images with the checklist above. Then revise the templates, not your standards.

Before you close the tab, confirm five things: the job sentence is written down, the aspect ratio matches the container, two variants exist, the winner is named with a date, and the bare link to Raphael AI is saved in your team note so the next person does not reinvent the workflow. That packing list is boring on purpose. Boring systems ship. When you need the generator again, open https://raphael.app, paste a template, and keep moving.

comments (0)

reviews (0)