Every small business owner knows this moment: you paid for a product shoot six months ago, and now your best-selling item comes in a new color. Or the hero photo background doesn’t match the season. Or your lead image still shows last year’s packaging. Studio work runs roughly $500–$3,000 per photographer day, or $50–$350 per finished image. That’s a real cost every time something changes. This month, Midjourney released its new Edit Model, tied to its V8.2 generator, and it changes that math in a practical way. Below I’ll walk you through what it does, how small businesses can use it for common jobs like background swaps and packaging updates, and the free open-source route through ComfyUI that delivers most of the same capability without a monthly subscription.
What Midjourney’s Edit Model Actually Does
Midjourney’s editing story used to be a patchwork. The web Editor ran on an older V6.1 backbone even while image generation had moved to V8. Omni Reference and Character Reference were locked to V7. If you wanted to change a background, swap an element, and expand a canvas on one image, you were juggling three model versions in a single workflow. That sent a lot of people back to Photoshop’s generative fill instead.
The V8.2 Edit Model pulls all of that into one unified tool. According to Midjourney’s official Edit Model documentation, it works three ways:
- Instruction editing — describe the change in plain language, like “make this product sit on a marble countertop,” instead of writing a full scene prompt.
- Reference-based generation — attach up to four reference images and blend them into a new composition, replacing both Omni Reference and Character Reference on V8.X.
- Inpainting and outpainting — brush over a region to change just that region, or expand the canvas beyond the original frame.
It also works with everything else in the V8 stack: style references, moodboards, personalization profiles, image prompts, and the –hd flag. In Discord, trigger it by appending –edit to your prompt and pasting your image URL. On the web, drag an image into the Imagine bar or use “Quick Edit” on any image you’ve already generated.
One honest caution from Midjourney’s own docs: this is an exploratory test. Even they acknowledge they don’t fully know the model’s capability ceiling yet. Keep that in mind as you experiment.
Four Small Business Workflows That Pay for Themselves
The workflows worth your time are the ones that replace a real, recurring expense. Here are the four I’d point you toward first.
1. The seasonal background swap. Your candle photo was shot against a dark backdrop in November. Prompt: put this candle on a bright kitchen windowsill with morning light –edit –ar 3:2 –raw. The –raw flag matters here — it dials down Midjourney’s default stylization so your instruction text dominates, keeping the product itself untouched while the environment changes. Always include –ar explicitly; the docs note the Edit model matches the aspect ratio of the first uploaded image otherwise, not your saved defaults.
2. The perspective shot you never took. You have a strong three-quarter view of your product but your listing page needs a straight-on shot. Prompt: let’s see this image from the front –edit. The docs confirm the model can rotate subjects to new camera angles from short instructions. It won’t be pixel-perfect to the physical object, but for social posts and ad creative it’s often indistinguishable from a reshoot.
3. The multi-product composite. Bundle shots used to require physically arranging products under studio lights. With the Edit Model, attach up to four reference images: a gift basket containing this soap, this lotion, and this candle on a rustic wood table –edit. A tip straight from the documentation: if attributes bleed between products, combine them into a single reference image before uploading.
4. The style conversion. Turn a product photo into a consistent branded illustration for social: hand drawn pencil sketch of this image –sref 2708276657 –edit. Midjourney recommends describing the style in words alongside the style reference to activate it more consistently.
The Free Route: Open-Source Image Editing in ComfyUI
Midjourney’s Edit Model starts at $10 a month for its Basic plan — you’ll want the $30 Standard plan if you’re editing in volume. There’s a zero-dollar alternative that runs on your own hardware: instruction-based image editing models inside ComfyUI, a free node-based workflow tool I’ve covered here before.
The standout is Qwen-Image-Edit, an open-weights editing model from Alibaba’s Qwen team. Thunder Compute’s 2026 guide calls it one of the most capable openly available models for instruction-based editing. It runs in ComfyUI with community workflow support, accepts the same kind of plain-language edit instructions as Midjourney’s Edit Model, and because the surrounding ecosystem is Apache-licensed, there are no per-image fees — ever.
A basic workflow looks like this:
- Install ComfyUI (free; runs on a consumer GPU with 16GB+ VRAM comfortably, less with quantized versions).
- Load a Qwen-Image-Edit workflow from ComfyUI’s built-in template library.
- Drop your product photo into the Load Image node.
- Type your edit instruction — “replace the background with a sunlit workshop interior” — into the prompt node.
- Generate, compare, and adjust. Edits run in seconds to roughly a minute depending on your GPU.
Community comparisons generally find Qwen-Image-Edit’s instruction-following slightly ahead of FLUX.1 Kontext, the other major open editing model — though Kontext tends to hold photographic fidelity better on e-commerce product shots, while Qwen leans slightly stylized. That split is practically useful: Qwen for ad creative and social graphics, Kontext when the photo needs to look like an unmanipulated photograph.
Pair either one with Z-Image Turbo — an Apache-2.0 licensed generation model that renders a 1024px image in roughly 2–3 seconds in ComfyUI — and you have a complete free studio: generate new scenes with Z-Image, edit and composite existing photos with Qwen-Image-Edit or Kontext. I covered the fundamentals of getting ComfyUI running in our guide to free AI image generation with Z-Image and ComfyUI — start there if the free route appeals to you.
Midjourney vs. ComfyUI: A Straight Cost Comparison
Here’s my honest read: if you edit a handful of images a month and want it done in five minutes, Midjourney’s Edit Model is the better purchase at $10. If you’re editing product catalogs weekly, the free stack pays for its learning curve within the first month. And nothing about it can be priced up or discontinued on someone else’s schedule.
What the cost comparison doesn’t capture is workflow. The businesses that get the most out of these tools aren’t the ones with the biggest AI budget. They’re the ones who define a repeatable process and stick to it — prompt templates, style references locked to their brand, a folder structure their whole team can follow.
Where AI Image Editing Fits — and Where It Doesn’t
For a long time, “update the marketing imagery” meant hiring someone, scheduling a shoot, and paying per asset. These editing tools compress that into minutes. But that efficiency still needs professional judgment attached to it.
Three practical rules before you dive in: keep your original photos untouched and always edit a copy. Verify each model’s current license before using outputs in commercial advertising. And be honest with customers if an AI-composited scene implies something the product isn’t — a marble countertop as context is fine; implying a feature your product doesn’t have is not. These aren’t cautions unique to AI. They’re the same standards any responsible visual approach applies to composite photography.
What’s the photo bottleneck in your business right now — the reshoot you keep postponing, or the image library that no longer matches your current branding? Drop a note in the comments. I’d genuinely like to hear what you’re running into.
Frequently Asked Questions
Do I need a paid Midjourney plan to use the Edit Model?
Yes. Midjourney requires a paid subscription for commercial use of generated or edited images. The Edit Model works with V8.1 and V8.2 on any current paid plan tier.
Is Qwen-Image-Edit really free for commercial work?
The model weights are openly published and the surrounding ecosystem is largely permissive-licensed. Always verify the current license for the specific checkpoint you download before using outputs in paid advertising — license terms on open-weights models do occasionally update.
Will AI image editing replace my product photographer?
Not for hero shots where accuracy to the real product matters. Legal standards and marketplace policies care that the photo actually represents the item. AI editing is strongest for backgrounds, composites, seasonal variations, and social creative built from photos you already own. The original shoot still earns its cost. What changes is how far that shoot can stretch.
How does Midjourney’s Edit Model compare to Photoshop’s generative fill?
Photoshop’s generative fill is stronger for precision masking and working inside an existing design file. Midjourney’s Edit Model is faster for plain-language instruction edits and compositing across multiple reference images. If you’re already in Photoshop for final output, generative fill makes sense. If you just need a quick background swap or a seasonal variant, Midjourney gets you there with less setup.
What GPU do I need to run ComfyUI with Qwen-Image-Edit?
You’ll want at least 16GB of VRAM for comfortable performance at full precision. Quantized versions of the model can run on 8GB cards with slower generation times. If you don’t have a capable GPU, cloud-based ComfyUI options exist — though at that point, the cost difference from Midjourney narrows considerably.
One thing I’ll be watching over the next few months: how well these editing models hold up as businesses start using them at catalog scale — dozens or hundreds of SKUs, not just a handful of hero shots. That’s where repeatable workflow structure really starts to matter. I’ll share what I’m seeing as it develops.
