You write the topic, choose whether you want 3, 4, or 5 slides, and the entire set is generated: a hook that invites swiping, one idea per slide, and a closing statement that encourages saving. All with the same body type size and text placed where the navigation dots do not cover it.
TikTok's photo mode does not forgive two things: that the first slide does not entice swiping, and that the navigation dots eat up the last line of text. This feature resolves both before you download anything.
You write the topic of the carousel in one sentence, choose how many slides you want —from 3 to 5, four by default— and select a style from the five in the catalog. From there, the app writes the content for the entire set in one go and formats it into vertical slides of 1080x1920 px.
The structure is not decided by the AI model: it is imposed by the template. Slide 1 is always the hook, with a short headline and a phrase that invites you to continue; the middle slides each carry one concrete idea, neither two nor half; and the last one is always the closing that asks to save the carousel. This way, a set never comes out with two hooks, three closings, or none at all. Each slide also has its discreet number (1/4, 2/4…) within the safe area, and only the first slide carries the swipe activator.
The measurements are what make this work. The photo mode reserves 200 px at the top, 140 px on the right for the button column, and 430 px at the bottom, because there is not only the description block: there is also the row of navigation dots for the carousel, which does not appear in a regular video. The text block is centered in the strip that remains free both in the feed and in the 3:4 window of the profile grid. That lower band is marked in the app itself as an estimate derived from the feed measurements, not as a measured data on phones: you will see it written on the screen.
The body type size is common to the entire set. The app measures the text of the five slides, looks for the smallest size needed by any of them, and applies it to all. Some utilization is lost in the short slides, yes, but the set reads as a family and not as five separate images. For the same reason, the dark veil over the background is identical in all. And the background is ONE for the entire set, with a luminance variation of ±3% per slide: sisters, not clones. This also means that asking for five slides does not cost five times more than asking for three.
Use it free without an account. If you want to save your result, you can sign up afterwards.
Save / Sign upOne idea per slide and the roles imposed by the template
the server assigns the hook to the first, the save closing to the last, and value to the middle ones, regardless of the model's response. A carousel never comes out with two hooks or without a closing.
The text falls where the navigation dots do not cover it
the safe area of photo mode reserves 430 px at the bottom —description block PLUS the row of carousel dots— and the block is centered on what remains free both in the feed and in the 3:4 window of the profile.
A body type for the entire set
the text of all the slides is measured and the size needed by the tightest one is applied to all. The five slides read as a series, not as five images made at different times.
Ventaja 4
A single background for the entire set, with ±3% luminance variation per slide. Requesting five slides does not multiply the generation cost by five, and the set maintains visual unity.
Touching a slide does not require redoing the set
you change its title or text and only that one is recomposed, over the same background, without using any AI generation. The title allows up to 70 characters and the body up to 170.
The row of navigation dots tells you where the problem is without scrolling through all
each dot shows its role (Hook, Value, Save) below it and turns red if the text of that slide does not fit or clashes with the interface. If everything fits, the stamp 'Coherent set · texts in safe zone' appears.
The AI does not write a single letter inside the image
it only paints the background and drafts the text. Each title, each body, and each slide number is composed by the server in vector format and printed after measuring them.
It must be uploaded in order or it won't upload correctly
the ZIP contains the slides numbered in the name (…-01.png, …-02.png) and a README with instructions; on mobile, the guided saving puts them into the camera roll one by one and in order.
A free account for ten seconds for the AI part. It does not connect to your TikTok account and does not ask for your password.
In the features that use AI, responses are generated by artificial intelligence and are not a substitute for professional advice.