🧰 The 2026 AI image-editing playbook: 4 jobs, 3 tools
Welcome back to AI Marketing News!
How many product photos on your phone have the wrong background?
The fix used to be an hour in Photoshop or hiring a freelancer.
In the last 8 days, three companies shipped tools for exactly that job. Google Pics launched on Sep 1st (Google). Midjourney's V8.2 edit model got a new Lightbox editor on Sep 3rd (Midjourney changelog). And ChatGPT Images 2.5 arrived on Sep 8th for every ChatGPT plan, including free (OpenAI).
All three do the same thing. They change a photo you already have.
The AI images that end up in the trash are the ones generated from nothing. Editing your own photo keeps the real product, the real face and the real storefront. AI only gets the boring part.
Below: the 4 image jobs most of you have every week, the tool that fits each one, and where all three still fall short. 👇
Social media has plenty of predictions. We want the reality.
What's actually working? What's getting harder? And what are social teams quietly dealing with behind the dashboards?
We're building the State of Social Report 2026 to find out, using answers from the people actually doing the work.
Give us 8 anonymous minutes and we'll turn the responses into a report on what's really happening across social media in 2026, and share the findings back with you.
Plus, you could walk away with a $250 gift card for your time.
📸 Play #1 — Product shots: ChatGPT Images 2.5
The job: a phone photo of your product that needs a clean background or better lighting for your website, Amazon listing, or IG post.
Why this one: OpenAI says Images 2.5 is "better at preserving subjects in reference photos." It also produces more natural lighting and richer textures, and handles transparent backgrounds better.
The move:
Upload — the real photo
Ask — describe the background or lighting you want, in plain words
Fix — use Image Comments to pin the one spot that came back wrong, instead of re-prompting the whole image
Repeat — OpenAI says 2.5 "follows editing instructions more reliably across multiple turns," so keep going in the same thread until it's right
Price: it's on every ChatGPT tier, including the free plan, on desktop, mobile and web.
🎯 Use it for: any image where the real product has to stay the real product.
⚠️ Watch for: the free plan has usage limits, and OpenAI doesn't publish the number. The speed claim is "up to 50%" faster than Images 2.0, so don't expect the demo speed every time.
—
🖼️ Play #2 — 10 versions of one post graphic: ChatGPT Images 2.5 again
The job: one promo graphic that needs a version per offer, per week, or per platform… without rebuilding it from scratch every time.
Why this one: 3 of the 4 new features in this release help when you're repeating a graphic.
Templates — pre-built formats for common creative types (like posters or merch)
Sketch — draw where the text and product go before you generate anything. Type @Sketch to open it
Prompt Sharing — share the prompt so a client, a VA or your other account can recreate the same graphic with their own photo
The move: build the master once from a Sketch, save the prompt, share it, and swap the photo each week.
🎯 Use it for: a repeatable weekly graphic.
—
🌍 Play #3 — Resize, translate and upscale one creative: Google Pics
The job: the same creative in 5 formats, in 2 languages, and a bigger version for print.
Why this one: Google Pics runs on Google's Nano Banana model. It lets you select one object or one text element in an image and change it without touching the rest. It translates the text inside the image, crops for web, social or print, and can "upscale your image to 2K or 4K" (Google Workspace blog). It runs several edits at once with an easy revert. And it lives at pics.new and inside Google Docs and Slides, so the flyer or deck you're fixing is already there.
Google's own pitch for it is creating and translating social and web assets for global markets.
The move:
Open — pics.new, or from inside Docs or Slides
Select — the object or text element you want changed
Edit — change, reformat or translate it
Crop — pick the format
Upscale — 2K or 4K if it's going to print
Price and plan: included with Google AI Pro or Ultra, or a Workspace Business Standard plan and up. Google lists no extra charge.
🎯 Use it for: localizing and resizing something you've already made.
⚠️ Watch for: there's no free tier and no Business Starter access. Editing straight from Drive is still "in the coming weeks."
—
🎨 Play #4 — The hero image with your brand's look: Midjourney V8.2
The job: the one campaign image or cover that has to look designed, in a consistent style, from your own reference images.
Why this one: Midjourney's V8.2 edit model takes plain-language instructions and handles inpainting and outpainting. It accepts up to 4 image references in one prompt alongside your style refs and moodboards (Midjourney). On Sep 3rd it got a new editor built into the lightbox, so every edit from a session sits in one place. There's also an experimental Change Style button.
The move: attach up to 4 references (your product, your palette, a past post you liked). Edit with plain instructions. Use the lightbox edit button to fix one region without redoing the whole image.
Price: Basic is $10/mo for 3.3 fast GPU hours, then Standard $30/mo, Pro $60/mo and Mega $120/mo, with 20% off annual (Midjourney docs). No free plan is mentioned.
🎯 Use it for: the 1-2 images a month that carry a campaign.
⚠️ Watch for: Midjourney's own words on this release: "there's gonna be a lot of edge cases." It also says moodboards and style refs "may work best with some extra prompt direction." And the lightbox editor lives on alpha.midjourney.com for now.
Built for Social Media Teams, Trained on What Performs in the Feed
Archie turns your webinars, articles, and decks into on-brand social posts for LinkedIn, Instagram, and Facebook, then recommends the best times to publish based on your own engagement data.
🧹 Check these 4 things before anything goes out
Every one of those plays ends with an export. Small text, hands, logos and exact brand colors are the 4 things to check by eye on every export, whichever tool made it.
Skipping that check is what the freelance market is now billing for. Freelancer.com listings for cleaning up AI output are up 87% (10,760 listings from Aug 2025 to Jun 2026). Upwork's AI-remediation gigs are up 70% year over year. Fiverr searches for "AI cleanup" are up 20x since 2023, and graphic design is the most common category (Search Engine Journal).
Freelancer.com CEO Matt Barrie says the cleanup often takes as long as creating from scratch, while clients budget it as quick and cheap. One client offered $500 to fix 13-15 AI illustrations and expected 15 minutes each. The illustrator, Todd Van Linda, charges $65/hour and turned it down.
My read: most of that cleanup work seems to come from images generated from nothing. Editing a real photo keeps you out of most of it, and the 4-point check keeps you out of the rest. None of the three replaces a designer for a brand system. All three replace the 20-minute Photoshop chore.
—
🧰 Cheat sheet
Product shot cleanup or new background: ChatGPT Images 2.5. Free plan works
Weekly post graphic, many versions: ChatGPT Images 2.5 with Templates and Prompt Sharing. Free plan works
Resize, translate, upscale to print: Google Pics. Google AI Pro/Ultra or Workspace Business Standard and up
Hero image in a consistent style: Midjourney V8.2 edit. From $10/mo
Every export: Check text, hands, logos and brand colors before it goes anywhere
—
☝️ The one-sentence version
Edit the photos you already have instead of generating from nothing. Start with the free ChatGPT tool for product shots and post graphics. Pay for Google Pics or Midjourney only when a job needs print sizing or a designed hero image.
Hit reply and tell me: which of these 4 jobs eats the most of your week?
—
📚 Sources
That’s all for today. Thanks for reading. Now…
Go BIG or go home!
~ Josh from “AI Marketing News”
Disclaimer: Some links may be affiliate links that pay us commissions.
Was this newsletter forwarded to you?
