Your whole episode at a glance.
Every panel of your script becomes a column: its shape, its characters, its prompt, its dialogue and its picture. Pick the panels you want and generate them in one go.
ProductsSections
38 links
Comic production / 01
Anyone can ask an AI for one good picture. The trouble starts at the second. A book needs three hundred, in which the same person keeps the same face, the same room keeps the same window, and every panel carries on from where the last one stopped. You direct. The application holds the order, the continuity and the count — and spends nothing without asking first.
Read this first
This is a production manager, not a magic button. It holds the order, the continuity and the count for a book of three hundred panels — but the drawing is done by AI services you pay for yourself, and the script has to exist before the first panel can be written. Three things have to be true before you start.
It turns each panel's notes into an image prompt. Anthropic Claude, OpenAI, Google or another from the list; you paste your own key and you pay that company directly, at their prices. The cheapest model is enough for this job — the application recommends it — and a whole episode of prompts costs cents, not euros. Sentiora resells nothing and adds nothing on top.
Prompts can also be built from a template, with no paid call at all. They come out weaker, and the reason is simple: a template cannot read the panel before this one. That is what the paid call buys.
This is the one that costs. Every panel is an image, and images are billed by the piece. The easiest route needs no key at all: you press a button, sign in to your own OpenArt account like on any website, and the panels come out of your subscription's credits. The other three providers want a developer account with money in it and a key pasted by hand.
In the screens below, thirteen panels are quoted at 13 × 60 credits ≈ 780 credits, and a background at 60. The application shows the estimate before it spends anything, and asks again above a limit you set.
The application reads a director's script; it does not write one. Turning a novel into episodes of seventy-five to a hundred panels is its own job, and it is done in Claude Code with the skill that comes with the application: you put the folder next to the book, open Claude Code there, and say make the script. It measures the book, asks how many episodes you want, writes a plan, then produces the episodes one at a time.
You can also write the script by hand, or bring one you already have — the format is documented and there is a full working example in the folder.
None of this is hidden in the application either: the panel that asks for money says what it will cost, the script check runs for free the moment you paste, and a batch stops before the first paid call if a character has no identity yet.
You stay in charge at every step: nothing is drawn until you say so, nothing is paid for without your yes, and every word sent to the AI is yours to read and change.








Paste a script, choose a look, and watch your story turn into pictures — panel by panel, with you in control.
Every panel of your script becomes a column: its shape, its characters, its prompt, its dialogue and its picture. Pick the panels you want and generate them in one go.
Heavy ink, clear line, noir, retro print, watercolour, manga and more — ready-made styles tested on real images. Add simple rules like “the same face in every panel” and “no text in the picture”.
Each location in your story is described once and reused in every scene set there, so a harbour stays the same harbour from the first page to the last.
Only the images you picked, large, in reading order, each with the lines that go over it. One click sends the episode to ComicBubble for lettering.
Work on several books, keep versions, and tidy up unused images whenever you like.
A closer look at what you work with every day.



Everything here follows from one decision: that the person paying for the images should be able to see, beforehand, what is about to be made, what it will be made from, and what is wrong with it.
This is meant for the real size of the job — hundreds of panels in an episode, not a dozen trial pictures. A script of 275 panels opens and comes apart in a moment, and the whole book sits in front of you, panel beside panel, so you can see at any moment what is finished and what is missing.
Every press that costs something says first how many images it will produce and what it is expected to come to. Every window holding work you have paid for asks before it closes.
You put your own provider keys into Settings, and the provider bills you directly for what you use. The application never stands between you and that account: your key, your account, their invoice.
Press What gets sent on any panel and read, word for word, the text that is about to go to the model. Each of the helpers — the previous panels, the position in the story, your notes — has its own switch, and off means the text genuinely does not travel, not that it has been hidden from the screen.
Dialogue and captions are deliberately kept out of the image, where they would come back drawn as crooked lettering underneath the real bubbles. They travel separately, in the lettering file.
WaveSpeed, Higgsfield and Pixazo for pictures; Anthropic, OpenAI and OpenRouter for the text models that write the prompts. Choose by price and by look, and change your mind as you go. Without a text key at all the application still writes prompts, from templates, for nothing — and pictures made somewhere else can be brought in at any point.
It runs on your own computer. The pictures, the projects and the keys stay in your own folders. Nothing is uploaded anywhere except the requests you make yourself, straight to the provider you chose.
English, Română, Español, Français, Deutsch, Português, Italiano, Русский, 中文, हिन्दी, বাংলা, 日本語, 한국어 and العربية — picked from one list, and the whole application follows.
Together they take you from a written story to a printed book: Comic Pipeline makes the pictures, ComicBubble adds the words, Panel Forge builds the pages. Each one passes your work to the next with a single click. But every program also works completely on its own — already have your art? Start with ComicBubble. Already lettered? Go straight to Panel Forge. You only buy what you need.
Paste your script or book. It plans every panel, keeps your characters looking the same, and creates the images with the AI service you choose.
586 hand-drawn balloons in 15 styles and a dozen comic fonts. Your dialogue lands on each panel, beside the right character.
Automatic page layouts, 708 drawn sound effects in 10 packs, 299 graphic shapes and 56 title styles. Export PNG, PDF or CBZ.
The application does not sell you generation and does not resell anybody's images. You open an account with an image provider, put the key into Settings, and pay that provider for what you use — nothing about the money passes through here. Without a text-model key it still writes every prompt, from templates, at no cost. What you get for your part of the work is the thing that is genuinely hard: three hundred panels that look like they belong to the same book.
Under construction
Sentiora is being built. Nothing here is finished, prices are not set, and no product can be downloaded yet.
Come back soon — or write to contact@sentiora.dev.