AI for Creatives — Week of July 3, 2026
Adobe buys Topaz, and a free gift from me.
Every week I take a look at some of the tools and ideas emerging in AI. And I focus on the things that are important and useful for creatives and entrepreneurs. No insider valuations or who is doing what… just stuff you can use today. A free gift for July 4th included. And yes, I use Claude CoWork to help compile this list.
Adobe just bought the plugin half of you already own
Adobe agreed to acquire Topaz Labs, the company behind Gigapixel, DeNoise, and the other AI upscaling and sharpening tools a lot of photographers already run. The plan is to fold Topaz’s on-device models directly into Photoshop, Lightroom, and Premiere instead of leaving them as a separate app you tab out to.
That round trip - export a noisy high-ISO file, run it through Topaz, bring it back in - is what’s going away.
What happens to the standalone Topaz apps and their pricing after the deal closes isn’t clear yet.
Nothing to do today except watch the rollout. But if you’re paying for both Topaz and Creative Cloud, this is worth tracking the next time either subscription renews.
(via The Neuron)
(So Adobe bought the IP and will roll it into their existing product. I love Gigapixel, so this means I don’t have to pay twice.)
Hourly billing is losing, and getting there first is the advantage
Consulting and professional-services firms are backing away from hourly billing because AI makes parts of the work faster and harder to meter by the hour, and clients want fees tied to results instead. Deloitte reportedly showed its own consultants’ data suggesting labor-based consulting could shrink sharply as a share of the market by 2035. McKinsey says more than 30% of its global fees are already tied to outcomes rather than hours.
The trap in hourly billing is obvious once AI enters the picture: the faster you get, the less you can bill. A fixed-fee or outcome-based package captures the value of the work instead of the time it took, which matters a lot once AI is doing your first-pass retouching or your rough cut.
The practical move: name the outcome you’re delivering, define the deliverable, and separate what AI speeds up from what still needs your eye. Price the package, not the hour.
(via The Neuron)
(So many of us have been saying this for years now. Bid and charge for the value, not the process. The ‘hourly’ worker will be more and more problematic for clients as they try to marry outcome to process… “if it takes longer it’s worth more.”
We do not want to play that game, folks.)
A free video editor that takes direction instead of clicks
Palmier Pro is a free video editor that connects to your Anthropic account through an API key. Drag in raw footage, a five-minute screen recording, a talking-head clip, product demo footage, and tell it in plain language what you want: “edit this down to about a minute, keep the strongest moments, cut the dead air.” Then iterate on captions, pacing, color, and crop before you export. It can also run as an MCP server inside Claude Code for a more automated pipeline.
This is aimed at the part of editing everyone dreads, the hours of scrubbing through footage looking for the good ten seconds. If you shoot video alongside stills, or you’re cutting client interviews and behind-the-scenes footage, this replaces a chunk of that scrubbing with a sentence.
(via The Rundown)
(OK, so I tried it. Put in some iPhone footage, some motion graphics, and random stuff I made on the fly. It was interesting, but you can almost hear your tokens being burned. If you are on the $20 a month plan, this thing will take you through your daily allotment fast. Real fast.)
Google built its new image and video models to work as a pair
Google’s Nano Banana 2 Lite generates a still image in about four seconds for roughly three and a half cents. Its companion, Gemini Omni Flash, generates and edits ten-second video clips for ten cents a second, and trails only ByteDance’s Seedance 2.0 on video-editing benchmarks. The two are built to chain: rough out a look as a still with Lite, then animate it with Omni Flash.
At Cannes Lions 2026, Kling AI and Higgsfield AI premiered an AI-generated campaign film for WEC and Le Mans, built with a hybrid of human direction, CGI, and AI generation, and multi-subject motion control that held up at commercial 4K output. Kling says it’s past 100 million users and 50,000 enterprise clients at this point.
For anyone pitching a director or a client on a look before committing to a real shoot day, that stills-to-motion loop is now fast and cheap enough to use as a pitch tool, not just a novelty.
(via The Rundown and The Neuron)
(I like the idea of creating an animated storyboard while pitching clients for motion, and larger video gigs. This can be a very powerful tool for getting something fast, and still having it look great.)
Keep client files off the cloud by running Claude Code against your own machine
Excellent AI Prompts ran a full setup guide for running Claude Code entirely offline against a local Ollama model instead of Anthropic’s servers, so nothing you feed it leaves your machine. Install Ollama 0.14.0 or newer, pick a model sized to your RAM (8GB up through 32GB and beyond changes what’s realistic to run), and raise the default context window, since Ollama’s 4096-token default is small enough to silently truncate a long session without telling you. Point ~/.claude/settings.json at localhost and you’re running local. The guide also includes a way to verify nothing actually left the machine, using a hosts-file block plus tcpdump.
If you’re handling unreleased client photos, signed contracts, or anything under NDA and you want AI help without those files touching a cloud server, this is the setup. Run local for the daily work and save the cloud model for the genuinely hard problems.
(via Excellent AI Prompts)
(Yeah, I see a little Mac Mini in my future. I am currently running Open Design on my Mac Studio, and while it is slow, it is definitely a glimpse into the future. Open Design is basically Claude Code running totally on your own device.)
Prompt of the Week
This one came out of a set of prompts framed around ADHD freelancers, but it’s just good pricing hygiene for anyone who freelances a camera, a keyboard, or a design file. It catches the two things that quietly wreck freelance income: pricing too low and scoping too loose.
You are a pricing and scope guard protecting a freelancer from undercharging and over-promising. [Describe the project you’re about to price, and what you were tempted to throw in for free.] Tell me...
Drop in the project you’re about to quote and what you were about to give away for free, and let it push back before you send the quote, not after you’ve already delivered the extra round of revisions.
If you’d rather talk than type, both this week’s Neuron and a piece from Ruben mentioned Wispr Flow, a dictation tool that turns spoken voice into clean, punctuated text across Mac, Windows, iPhone, and Android. Ruben says speaking his prompts instead of typing them roughly quadrupled his speed and got him richer, more specific prompts in the process. Worth trying if typing out a long prompt is the thing that keeps you from using one.
(via Excellent AI Prompts)
Get a free month of Wisprflow on me.
Happy Fourth of July to everyone (even if you aren’t in the US, you have yourself a wonderful Saturday.)
To celebrate this day, I’m giving you this handy little tool to have some fun with.
The Landing Page Builder.
Link fixed. Sorry.
Insert your Claude API Key, and build a custom landing page in one swoop. (Instructions on getting an APi Key are in the tool.)
Note: I suggest you have a decent Design.md file to keep it from looking too generic. I will have more on building a good design.md file this coming week.
For power users, run the final HTML through Impeccable or Huashua for a cleaner page.
Download The Landing Page Builder here.
IMAGES
Your images go into a folder next to the HTML file named “images”.
Follow this strict naming and size convention:
header.jpg (1600 x 1080)
footer.jpg (1600 x 450
solution.jpg (900x600)
cta.jpg (900x600)
For gallery images: 8 images in portrait format (500x750)
gal1.jpg, gal2.jpg, etc. for a total of 8.
If you wish to change any image on the site, just drop a new one in the images folder with the same name. Done.
Example (using a design.md file)
If you’re over 40 and still hungry to make, build, and create, stick around. This space is for people who aren’t done yet (and never will be). I’ve got five decades of wins, failures, comebacks, and creative battles under my belt, and I’m sharing everything that still works—and burning the rest.
No fluff.
No hustle porn.
Just real tools for building a creative life on your own damn terms.
My Personal Website
Me on Medium.