EP 017 · 58 MIN ADDING TO AN IMAGE IS HARDER THAN REMOVING
Cracking Midjourney's New Editor Features (Retexture & External Editor)
The short version
Episode 17 puts Midjourney's newly launched external Editor and Retexture (depth ControlNet) features through their paces about a week after release. Rory Flynn and Drew Brucker test real client use cases, like reskinning a décor product into new room styles, and find that removing or replacing elements works well while adding new ones, such as an end table, mostly fails unless a separate generated image is fed back in as an image reference.
Key takeaways
- 01
Midjourney's new Editor and Retexture features are gated to 10,000+ image generations or 12+ months on an annual plan, and both hosts call it clearly beta.
- 02
Retexture lets you drop an existing, client-approved product photo into new styles while keeping the same camera angle and lighting — reshoots become prompt swaps.
- 03
Removing, replacing or restyling an element in the editor works well; adding something new, like a rug or an end table, mostly fails outright.
- 04
When adding a new object won't take, generate that object on its own first, then bring it back in as an image reference to force correct placement.
- 05
Uploading a photo from your own computer gets flagged by moderation far more often than pasting the same image's URL, in Rory's experience.
- 06
Retexture fidelity drops on faces, especially in the mid- or background, but holds up well on objects and foreground elements.
- 07
An old SREF-style code may not survive Midjourney's move to V7, but Editor and Retexture could let you apply that code onto a new V7 image afterward.
Receipts · said on the record
“You're utilizing this to then create all these amazing scenes through it. And that to me has been like one of these things where I'm just like, damn, this does change the game.”
“Getting things to add versus change were hard. To get things to change wasn't hard, or replace, or delete even, but to add was sort of the tricky part.”
“I just created an image of a rim that I was thinking, and I put it in there as an image reference. Then, bang, there it is.”
“The Photoshop generator is not as good as the mid-journey generator. So if you're trying to do larger details, you can tell there's a difference.”
“I think text prompting in general is kind of like a meta thought, but it's evolved, it's gotten easier in some ways … I think maybe text prompting in general is just a temporary thing. It's gonna eventually not be text at all.”
Chapters
Questions this episode answers
What are Midjourney's Editor and Retexture features?
Launched in late October 2024 as a beta gated to accounts with 10,000+ image generations or 12+ months on an annual plan, Editor lets you bring in an external image and edit it, while Retexture uses depth ControlNet to restyle an image's materials and surfaces while preserving its underlying structure.
Can Midjourney's editor add new objects to an existing image?
Not reliably: Rory Flynn and Drew Brucker found that removing, replacing or restyling an element worked well, but adding something entirely new, like an end table or a car spoiler, mostly failed by text prompt alone.
How do you force Midjourney's editor to add a specific object?
Generate the object on its own first, such as a wheel rim or an end table, then feed that image back in as an image reference inside the editor — Rory Flynn's workaround for the tool's weak text-based adding.
Is Midjourney's generative fill as good as Photoshop's?
Rory Flynn says Midjourney's generator produces better large-area detail than Photoshop's, but Photoshop remains better for precise, small local edits and removals, so neither fully replaces the other yet.
Will old Midjourney style-reference (SREF) codes still work after V7?
Drew Brucker and Rory Flynn speculate that even if SREF codes change on the new model, the Editor and Retexture tools could let you generate on the new version and then apply an old style code on top.
Show notes
Episode 17—Cracking Midjourney's Newest Features in 72 Hours (External & Retexture)
In this episode, Drew and Rory share their Halloween experiences, discussing costumes and childhood memories.
They break down the new external editor and retexture features, sharing their first-hand experiences as beta testers.
The duo dissects real client applications, particularly in interior design, where they've discovered powerful workflows for transforming existing spaces. They reveal crucial techniques for adding elements to images, demonstrating how image references can solve common editing challenges.
Through practical examples - from Tesla modifications to speech bubbles - they explore both the capabilities and current limitations of these new tools.
The conversation culminates with insights on text generation improvements, quality settings, and a glimpse into Midjourney's upcoming personalization features.
