ContentOS
Picture your ideas. Create a complete series of visuals and copy in one click.
Workflow diagram reconstructed from the project account.
One image can impress. A series has to finish the thought.
ContentOS is built around “Picture your ideas. Create a series of visuals and copy in one click.” The word I care about is “series.” One image can catch the eye; a sequence needs an order that helps it explain something.
The product entered formal operation in May 2026 and has close to 100 long-term unique users. As founder, I lead product strategy, design, project management and operations alongside an independent team. That figure describes long-term users, rather than daily active, monthly active or paying users.
I want someone to describe an idea once, choose the number of images, and receive a sequence organized around the topic. Explaining an idea should not require learning prompt engineering first.
Arrange the images like a small magazine
I think of multi-image creation as editing a small magazine. The cover attracts attention; later pages explain the point, add information and bring it to a close. If every page tries to be the cover, the result may feel lively without saying much.
I treat each image as an “information frame”: decide what it needs to say before working on style, composition and detail. The sequence also needs to be checked as a whole. Do two images repeat each other? Is a step missing? Do the image and its text agree?
The product organizes generation in that order. A director-style Agent creates an outline, divides the content and identifies the point of each part before generating image prompts and images. Each image has a job before generation begins.
How much of a tool do you need to change a headline?
We initially tried Canvas region selection so users could point to the place they wanted to change. We later found that text descriptions achieved similar editing precision in this use case, so we simplified the interface.
Users want to make the image look the way they need it to. If they can describe a change in words and the model can follow it, do they need another interaction to learn? More controls may allow finer edits, but they also add something to understand. I have to weigh the benefit against that effort.
Complex local edits may still need region tools. Removing some operations was a choice for this particular task. The time spent building a feature does not make it useful everywhere.
After generation, people need to revise and take the result away
ContentOS provides previews, copying, downloads, regeneration and history. A multimodal model assists with image and text checks. Users review the result and decide what to change and whether to generate again.
What I want to keep observing is how many revisions it takes before someone wants to use the sequence. A first image may be a pleasant surprise; starting over for every small change soon becomes frustrating. We do not yet have a consistent efficiency measure that answers this question.
“One click” refers to generation. Platform previews help with presentation checks; creators still verify facts and decide whether to publish. It does not mean automatic publishing to every platform. The real screenshot below retains the sample's text-rendering errors. People encounter those errors when using the product, and they still need work.
Next time, record why something needed to be redone
Building this product has made me care more about whether users understand the generation process and can continue editing when something goes wrong. Models and prompts will change. Who the content is for and what each image should explain still need to be clear.
In later iterations, I want to record revision reasons more precisely. Was the outline poorly divided? Did the text and image disagree? Were words rendered incorrectly? The next step might be an outline change, a prompt revision or a new image. At least the record should say more than “generated again.”
