EP 056 · 59 MIN FLUX FACE IS REAL. MIDJOURNEY STILL WINS.
Go Flux Yourself: Midjourney Still Runs the Realism Game
The short version
Episode 56 has Rory Flynn and Drew Brucker run Grok Imagine's infinite-scroll generator live, test Magnific's Mystic 3 against Midjourney on portraits and food, and land on a running bit, 'flux face,' after Flux-powered images keep looking too sterile next to Midjourney's lived-in realism. Rory also argues Midjourney remains the only tool that reliably outpaints and resizes images, beating Photoshop, Firefly and Flux at that one job.
Key takeaways
- 01
Midjourney is still the only tool that reliably outpaints and resizes images; Photoshop, Firefly and Flux all fail at extending a square into a different aspect ratio.
- 02
If you need brush-precision edits that Nano Banana or Flux Context can't do with text alone, paint a mask in Weavy's compositor node and prompt the bounding box.
- 03
Start prompts from Midjourney's Stills Archive instead of a blank box; Drew says it taps the training data harder and produces noticeably more cinematic, polished results.
- 04
Stop upscaling Midjourney images you want to look lo-fi or analog; sharpening kills the unpolished look, so lean on 'smart grain' and noise settings instead.
- 05
Mystic 3's new 'state of the art' 4K model nails skin pores, fingerprints and hand detail but still struggles with eyelash volume and multi-utensil scenes.
- 06
Right-click any parameter in Midjourney (aspect ratio, chaos value) to instantly filter your own library by it; hold Option on a grid to reveal upscale subtle/creative/loop.
- 07
Grok Imagine's output is recognizably 'flux face,' too sterile and background-light next to Midjourney; still useful as a fast way to harvest a style to reference elsewhere.
Receipts · said on the record
“It's the only tool that can outpaint accurately … if I have a square and I need to turn it into three by two, Photoshop doesn't do the trick, Firefly doesn't do the trick, Flux doesn't do the trick.”
“They see that and they're like, oh, that's why it's not real. But I guarantee I could show you stuff in my camera viewfinder that you would see and you're like, no, that's not real.”
“That's a little hack for anyone using Nano, C.Dream, or Flux Context — if you want to use a brush, you have to do it externally. I'll do it in the compositor or the painter node.”
“You can see the flux in here by the faces, right? There's a flux face that I like to call out a lot of times.”
“That might be on the thumbnail, ladies and gentlemen. Go flux yourself.”
Chapters
- 0:00 Intro, eyeballs, and a Friday brain check
- 2:05 Contact lens horror stories, Mission Impossible, Black Mirror eyes
- 3:07 Stranger Things Season 5 hype and binge vs weekly TV
- 4:51 Movies, biopics, sports docs, and couch season setting in
- 6:23 Cowboys documentary, sports pipelines, and TV as passive story feed
- 7:00 AI overload, nobody keeping up, and why this pod exists
- 8:30 Midjourney profiles, Style Creator, and new personalization talk
- 9:29 Like/dislike buttons as hidden training data and 7:3 aspect ratio love
- 10:35 Stills Archive, cinematic framing, and cleaner compositions
- 12:00 Style Explorer vs old-school SREF and what quietly vanished
- 13:16 Three under-the-radar Midjourney Smart Search + right-click + Option-upscale tweaks
- 15:35 V8, fewer wall-of-text prompts, and a move toward visual controls
- 18:12 First look at Grok Imagine’s interface and infinite scroll feel
- 19:35 Sora-style endless bottom feed, variants, and “make video” in Grok
- 22:51 Cinematic looks, color grading, and Grok as “idea and curate” engine
- 24:19 Live NSFW surprise inside Grok Imagine and instant rating change
- 25:23 Finding Grok history, stills, and video exports with sound
- 26:31 Who actually gets Grok video and Drew’s first real reaction to using it
- 27:38 Mystic 3 enters the chat and upscaling less for analog vibes
- 29:02 Why “too sharp” screams AI and how grain + smart detail saves realism
- 30:18 Outpainting, editing, and why Midjourney still wins surgical compositing
- 35:01 Mystic 3 V3 screen-share and first impressions
- 35:45 Editorial portraits, skin detail, eyelashes, and hands that finally look human
- 37:26 Mystic 3 model comparisons: Zen, State-of-the-Art, and weird description blur
- 39:16 Zooming all the way into pores, fingerprints, and micro skin texture
- 43:44 Cocktail and food prompts where Mystic falls behind Midjourney
- 50:05 Nano Banana 2 rumors, native 4K wishes, and how Midjourney might respond
- 50:58 Why Midjourney still rules lo-fi, disposable camera, and Polaroid-style shots
- 52:16 Grok Imagine vs Flux vs Midjourney for lived-in Y2K flash photos
- 53:39 Flux face, direct flash tests, and “go flux yourself” is born
- 55:30 Nodes, Grok workflows, and why scrolling is faster than wiring graphs
- 56:01 Why Midjourney is avoiding node-based interfaces on purpose
- 57:05 Final sendoff: go flux yourself and get out of here
Questions this episode answers
Why do Rory and Drew say Flux-powered images have 'flux face'?
Running the same Y2K flash-photo prompt on Grok Imagine (which uses Flux) came back too sterile and light on background detail, with a recognizable, generic look they nicknamed 'flux face,' versus Midjourney's grittier result.
Why is Midjourney still used for outpainting and resizing?
Rory says it's the only tool that reliably outpaints a square image into a new aspect ratio; Photoshop, Firefly and Flux all failed the same test for him.
How does Mystic 3 compare to Midjourney on realism?
Rory ran identical prompts in both; Mystic 3's new state-of-the-art 4K model impressed on skin, hand and food texture but still struggled with eyelash volume and multi-utensil scenes like forks and spoons.
What is Grok Imagine and is it free?
It's Grok's image tool, powered by models like Flux, with an infinite-scroll browsing feel similar to Sora; Rory and Drew found it free with limited usage quotas on the basic plan.
How do you get brush-precision edits in Nano Banana or Flux Context?
Rory says neither offers a native brush, so he paints a mask in Weavy's compositor or painter node and prompts changes to that specific bounding box instead.
Show notes
Drew and Rory start Fast Hours with eyeball horror, Stranger Things hype, and the idea of AI-powered contact lenses before stumbling straight into the real mind-melt: Midjourney, Grok Imagine, Mystic 3, and Flux all colliding in one episode.
They roast their own prompts, trigger an accidental NSF-DoubleU moment live inside Grok, argue about “flux face,” and still somehow manage to pull out real, practical tips for people trying to make better AI images without losing their minds.
Across an hour of chaos, they unpack Midjourney v8’s subtle shifts, hidden personalization signals, Style Explorer tricks, Smart Search shortcuts, Grok’s Sora-style infinite feed, Mystic 3’s scary-good skin detail, and why Midjourney still owns lo-fi, lived-in, “shot-on-a-phone” energy.
If you care about composition, cinematic ratios, editorial portraits, food realism, or just want to hear two people dunk on Flux and node editors while actually teaching you something, this one hits.
Listeners will come away knowing how to use stills archive for composition, when to skip upscales for more analog realism, how Grok Imagine’s image + video workflow really behaves, and where Mystic 3 can replace Midjourney in a serious portrait or product stack.
--
⏱️ Midjourney Fast Hour 0:00 Intro, eyeballs, and a Friday brain check 2:05 Contact lens horror stories, Mission Impossible, Black Mirror eyes 3:07 Stranger Things Season 5 hype and binge vs weekly TV 4:51 Movies, biopics, sports docs, and couch season setting in 6:23 Cowboys documentary, sports pipelines, and TV as passive story feed 7:00 AI overload, nobody keeping up, and why this pod exists 8:30 Midjourney profiles, Style Creator, and new personalization talk 9:29 Like/dislike buttons as hidden training data and 7:3 aspect ratio love 10:35 Stills Archive, cinematic framing, and cleaner compositions 12:00 Style Explorer vs old-school SREF and what quietly vanished 13:16 Three under-the-radar Midjourney Smart Search + right-click + Option-upscale tweaks 15:35 V8, fewer wall-of-text prompts, and a move toward visual controls 18:12 First look at Grok Imagine’s interface and infinite scroll feel 19:35 Sora-style endless bottom feed, variants, and “make video” in Grok 22:51 Cinematic looks, color grading, and Grok as “idea and curate” engine 24:19 Live NSFW surprise inside Grok Imagine and instant rating change 25:23 Finding Grok history, stills, and video exports with sound 26:31 Who actually gets Grok video and Drew’s first real reaction to using it 27:38 Mystic 3 enters the chat and upscaling less for analog vibes 29:02 Why “too sharp” screams AI and how grain + smart detail saves realism 30:18 Outpainting, editing, and why Midjourney still wins surgical compositing 35:01 Mystic 3 V3 screen-share and first impressions 35:45 Editorial portraits, skin detail, eyelashes, and hands that finally look human 37:26 Mystic 3 model comparisons: Zen, State-of-the-Art, and weird description blur 39:16 Zooming all the way into pores, fingerprints, and micro skin texture 43:44 Cocktail and food prompts where Mystic falls behind Midjourney 50:05 Nano Banana 2 rumors, native 4K wishes, and how Midjourney might respond 50:58 Why Midjourney still rules lo-fi, disposable camera, and Polaroid-style shots 52:16 Grok Imagine vs Flux vs Midjourney for lived-in Y2K flash photos 53:39 Flux face, direct flash tests, and “go flux yourself” is born 55:30 Nodes, Grok workflows, and why scrolling is faster than wiring graphs 56:01 Why Midjourney is avoiding node-based interfaces on purpose 57:05 Final sendoff: go flux yourself and get out of here
