How I AI — Episode Summary

Episode Title:

Mastering Midjourney: How to create consistent, beautiful brand imagery without complex prompts
Guest: Jamey Gannon (AI Creative Director)
Host: Claire Vo
Date: March 9, 2026

Overview

This episode dives deep into practical, repeatable workflows for building consistently beautiful and unique brand imagery using AI generative tools—primarily Midjourney and Nano Banana. Host Claire Vo and guest Jamey Gannon break down the exact process Jamey uses to create standout visual assets, emphasizing simple prompts, leveraging visual references, and strategies for maintaining stylistic cohesion across large image sets. The episode is packed with insights for designers, founders, and anyone wanting to make the most of modern image-generating AI.

Key Discussion Points & Insights

1. Why Consistency is the Real Challenge in AI Visuals

• Many can get a one-off, striking AI image, but it’s building a coherent “portfolio” or brand language that’s hard. Jamey’s years in Midjourney have been about developing a tight, manicured workflow so you’re not “pulling your hair out prompting all day.”
  • Quote: “It all comes down to having a very tight and manicured process…So you're not pulling your hair out prompting all day.” — Jamey (00:00)

2. Mood Boards & Visual Language for AI

• Jamey starts every project by crafting a mood board on Pinterest or Cosmos to establish the “vibe.” This is imported to Midjourney as a visual guide for the AI.

• Style References (SREFs): Instead of general mood boards, using select images as “style refs” gives Midjourney stronger, more specific direction on look and feel.

  • Quote: “A picture is worth a thousand words. Like literally, a picture to an LLM is worth a thousand words.” — Claire (00:12, 06:13)

• Pro Tip: If your generated images don’t match your mood board, use GPT or Claude to help analyze why—this builds your “creative vocabulary.”

3. The Creative & Iterative Process in Midjourney

• Start simple: Early prompts are broad, e.g., “beautiful female model” or “astronaut,” just to see what comes out. Don’t be too precious—generate fast to get baseline outputs.

• Iterate: Study results, tweak or swap out SREFs. If a style pulls too much of one color (e.g., too green), remove that reference. Intuition develops as you use the tool.

  • Quote: “I find that using srefs as the mood board instead...can give much better results...It's something that comes with intuition.” — Jamey (08:45)

• Test Prompts: Jamey uses recurring subjects that work well (cats, astronauts, runners) because of their rich training data and variety.

4. Personalization Codes (Advanced Consistency)

• Personalization Codes: Train Midjourney’s personalization model by curating image pairs, voting on your favorites, and skipping aggressively. This fine-tunes outputs to your exact taste (“late 2025 aesthetic”).
  • Quote: “When you’re creating a personalization code, Midjourney puts you through this endless flop matrix...you're just telling it what you like.” — Jamey (13:22)
• Combine these codes with style refs for even tighter, on-brand results.

5. Leveling Up Prompts (But Still Keeping it Simple)

• Name-Dropping as a Shortcut: Referencing magazines (“Dazed editorial photo shoot,” “Vogue”), artists, or high-fashion terms compresses complex visual guidance into a word or two.

  • Quote: “Mentioning like Vogue or high fashion or even like a different artist name is a great way to tell the model a ton of stuff without actually having to tell a ton of stuff.”—Jamey (00:25, 16:48)

• Image References: Use these to control composition. Tweaking by cropping can suppress unwanted elements (e.g., cropping out a prominent bubblegum in a photo).

• Camera Settings as Style Codes: Pasting in actual camera names/models (“Sony RX100”) instantly shifts aesthetics to more realistic or era-accurate looks.

  • Quote: “Cameras are cheat codes for styling.” — Claire (24:02)

• Literal Prompts: Don’t be afraid to just say what you want—“New York skyline can be seen in the window” or “on a matte black leather couch.”

6. Scaling Visual Systems & Delivering to Clients

• Once you have a tight prompt and SREF combination, you can create large, consistent image libraries. Jamey adds the final “successful” images back into new mood boards for further generations.

• Organizes and delivers image packages using Figma, including prompt/code documentation so clients can reproduce or iterate themselves.

  • Quote: “I’ll give clients this. I’ll give them a set of images…they can go do this for themselves.” — Jamey (33:41)

• This model is more collaborative and less dependent on expensive re-hires for every asset.

7. Editing & Upscaling: Nano Banana and Flora

• For detailed edits (fixing “AI hands,” adding/removing objects, upscaling), Jamey uses Nano Banana and Flora as “Photoshop you can talk to.”

• Nano Banana is especially powerful for upscaling and minor replacements; it understands reasoned, high-level commands (“replace computer with 2026 MacBook Pro,” “keep position/scale the same”).

  • Quote: “Nano Banana literally is just Photoshop. That's exactly how you should think of it. You're just able to speak to Photoshop essentially.” — Jamey (35:56, 36:51)

• Facial/Brand Consistency: Jamey uses personal reference selfies to generate realistic, brand-cohesive headshots and thumbnails.

8. Limitations & Biases in AI Models

• Discussion about demographic and contextual bias in training data: getting Midjourney to output certain demographic types/subjects can be hard due to biases ("It refuses to do Asian men, athletes default to muscular Black men, etc.").
  • Quote: “Every athlete, if you want an athlete in midjourney, it's going to be a very large black man...You cannot get a white athlete in printer.” — Jamey (42:18)

Notable Quotes & Memorable Moments

• On simplicity:
  “I try and avoid prompting at all costs in my process.” — Jamey (08:18)
• On shortcuts:
  “Everyone just wants to be lazier with their prompting. No one wants...these too long prompts.” — Claire (24:02)
• On creative block:
  “If AI’s not giving what we want—take a break, come back fresh. Often it's just about really honest, even brutal diagnosis: What exactly is wrong with these images?” — Jamey (47:17)
• On inspiration:
  “Pinterest is my home base...also I do a daily taste practice. Get really on top of saving and archiving stuff.” — Jamey (43:45)
• On working with clients:
  “You’re gonna value me for all this upfront work...and then now you can go do this for yourself. It’s a very different model of providing service.” — Claire (34:33)
• On the future of image work:
  “Midjourney was the first tool that just switched my mind about what was possible with AI. It's one of the more soulful AI experiences.” — Claire (26:51)

Step-by-Step Midjourney Workflow Cheat Sheet

1. Gather Inspiration (03:41)

   • Make a mood board (Pinterest, Cosmos) tailored to desired vibe.
   • Don’t overthink – broad concepts, visual juxtapositions.

2. Set Initial SREFs/References (06:13, 10:34)

   • Drag select images as SREFs into Midjourney.
   • Start with simple prompts (e.g., “astronaut,” “female model”).

3. Create & Iterate (12:40)

   • Generate lots of quick outputs; don’t be precious.
   • Diagnose mismatches—tweak SREFs, remove problem images.
   • Use simple test prompts for subjects rich in training data (cats, astronauts).

4. Fine-Tune with Personalization Codes (13:22)

   • Create personalization models by voting/skipping in Midjourney.
   • Skip aggressively—only keep what truly matches your brand style.

5. Prompt Strategically (16:48, 21:30)

   • Name-drop magazines or artists for instant style strength.
   • Add literal details only when needed (“New York skyline,” camera types).

6. Scale & Organize (29:01, 33:41)

   • Build new mood boards from successful outputs.
   • Mix and match mood boards/SREFs for new subjects.
   • Share package with clients via Figma, including prompt/codes for reuse.

7. Edit & Upscale as Needed (35:56, 37:55)

   • Use Nano Banana/Flora for Photoshop-style edits and upscaling.
   • Leverage personal photo references for facial consistency.

Lightning Round (43:11 – 47:17)

• Inspiration Sources: Maintain Twitter lists of aesthetic accounts, Tumblr-style feeds, and fashion/visual culture boards. Use Cosmos for art direction, Pinterest as home base.
• Daily Practice: Save and archive inspiration constantly; keep mood boards streamlined.
• Handling AI Frustration: Step away, come back fresh, diagnose with brutal honesty, trim SREFs/mood boards, and restart if needed.

Where to Find Jamey Gannon

• X/Twitter: @amiegannon / Tech Bimbo
• AI Course: “The AI Creative Director” on Maven
  (Details on Jamey’s X or website)
  “Meant to be able to make you create consistent client-level work like I showed today.” (48:36)

Episode Timestamps for Key Segments

• 00:00 - 03:16: The promise/challenge of consistent AI brand imagery
• 03:41 - 08:45: Mood boards, SREFs, and fast creative generation
• 08:45 - 14:32: Diagnosing mismatches, using SREFs, and personalization codes explained
• 16:48 - 24:58: Moving to advanced prompts, name-dropping, image/camera techniques, literal language
• 29:01 - 35:29: Scaling and packaging final image sets, client delivery models
• 35:56 - 43:11: Using Nano Banana/Flora for editing, upscaling, and personal branding
• 43:45 - 46:38: Sources of inspiration, archiving taste, building visual language
• 47:17 - 48:28: Overcoming creative blocks, honesty, workflow resets
• 48:36 - End: Where to find Jamey, course plug, wrap-up

Final Thoughts

Jamey’s approach demonstrates that consistent, on-brand, “beautiful and unique” AI imagery is the result of clever visual groundwork, smart short-cutting, and honest iteration—not convoluted or lengthy prompts. Whether you’re a creative director, marketer, or founder, these workflows make AI both more powerful and more approachable.