|
|
Was this forwarded to you? Get The Fourth Estate AI Brief in your inbox →
|
|
|
|
View this issue in your browser
Friday Deep-Dive
Now you can point at the one thing that is wrong
In August I ran one of our own real estate ads through ChatGPT and told you the honest result: about ninety percent right, and the last ten percent was maddening. You could see what was wrong. You just could not tell it that without it redrawing everything and breaking three other things. On Tuesday OpenAI shipped the fix. ChatGPT Images 2.5 lets you comment on one spot in an image and change only that, draw a rough layout by hand and have it finish the job, and start from a template instead of a blank box. It went out to every plan, free included. Below: the four new tools in plain terms, five things your team can do with them next week, and the step almost everybody skips, which is making the image actually work on your website.
|
ChatGPT's image tools just got usable for real work
OpenAI shipped ChatGPT Images 2.5 on Tuesday, and for once the interesting part is not the model. It is the four things you can put your hands on. You can now draw a rough layout inside ChatGPT and have it build on top of your sketch. You can start from a Poster or Merch template instead of a blank box. You can place a comment on one spot in an image, the way you would red-circle a proof, and change only that. And you can share the prompt along with the picture, so one person works out your house ad look and everybody else can match it. It is also up to fifty percent faster, and it is on every tier, free included. In the full post: the plain-English tour, five uses for a newspaper team, a copy-and-paste prompt, the four steps that turn a pretty picture into a web ready one, and the three rules that keep this out of trouble.
Most AI image news is a demo you watch once and forget. This one lands on real work you already do every week: the spec ad for the account that keeps saying no, the house ad filling a hole on page 6 at eleven at night, the column with no art. The editing change in particular is the thing that was standing between a promising toy and something a designer would actually keep open. Worth twenty minutes this weekend.
Source →
The full walkthrough, tool by tool →
|
|
What is?
C2PA (Content Credentials)
What it is: C2PA is a technical standard for attaching a tamper-evident history to a file. Think of it as a nutrition label baked into the image itself: what made it, when, and what was done to it since. It rides along invisibly in the file. OpenAI says every image ChatGPT makes carries C2PA data plus an invisible watermark, and Adobe, Google, and most major camera makers are in on the same standard. You read it with a checker such as the free one at contentcredentials.org/verify.
Why publishers care: Two reasons, and they point in opposite directions. First, anything your own staff generates is now stamped, whether or not you label it. That is one more argument for saying so in the caption yourself rather than having a reader discover it. Second, and more useful, this is becoming a real tool for the incoming photo. When a reader emails you a dramatic picture of a crash or a storm, dropping it into a Content Credentials checker takes about fifteen seconds and can tell you it came out of an AI tool. It is not a lie detector. Stripping the data is easy, and a missing credential proves nothing either way. But a credential that is present and says ChatGPT is about as clear an answer as you will get before deadline.
Visit contentcredentials.org →
|
|
|
|
Like what you see here, please forward to someone who you think would enjoy reading about AI in publishing!
|
|
Trevor
Trevor Slette
Co-founder, Quadd.ai · 28-year community publisher
trevors@quadd.ai
·
Book a 15-min call
·
LinkedIn
P.S. If you try the Sketch tool on a real ad this weekend, send me what you drew and what came back. I will put the best before-and-after in a future issue, with your paper's name on it if you want it there.
|
|
|
|
|
From meeting audio to usable copy.
Upload a council meeting, an interview, a coach's post-game scrum. Get a clean transcript and an AI summary, then ask Quadd to compose a first-draft story or paste your own prompt, and get formatted output in your publication's exact style. No more scrubbing through forty minutes of audio at 9pm because the lede quote was somewhere in the middle. Trained on the way reporters actually use audio, not the way podcast editors do.
See how transcription works →
|
|
|
