Frequently asked questions.
For the full pipeline explanation, see how it works.
Product
What does Castia do?
CASTIA helps adapt comic and webtoon dialogue while preserving character voice, emotional intent, and scene context. Each adapted line is shown with a plain-language reason for the creative choice.
How is this different from Google Translate or a single AI prompt?
Every line goes through a three-stage "Writers' Room": a Translator produces one literal anchor translation, a Creative Adapter produces five differently-angled rewrites, and a Judge scores each rewrite against that anchor and picks a winner — writing down the specific reason it departs from a literal translation. A single translation call has no anchor to check itself against and no record of why a word choice changed.
Which languages are supported?
The current engine supports English, Hindi, Japanese, Korean, Spanish, and Urdu. Available directions depend on the deployed language configuration.
What does CASTIA do for comics and webtoons?
Upload comic or webtoon panels to detect speech-bubble text, review the extracted dialogue, adapt lines with character and scene context in mind, and optionally redraw panels with the adapted text typeset in. Current limitations include sound-effect text over artwork not being redrawn and OCR reading order being imperfect for some layouts.
Pricing & access
Do I need an account?
No — a limited number of adaptations per day work with no account at all. Signing in with Google adds a history of your past adaptations and lets you star or collect results so they're there next time.
How much does it cost?
Castia is free to use right now, for everyone. Paid tiers are listed on the pricing page as what's being built toward, but billing isn't live yet — nothing is charged today.
Technical
Is there an API?
Yes, currently in Beta: POST /v1/comics/adapt for comic panels, authenticated with an API key you generate from your dashboard settings, with a daily request limit per key.
What happens to the dialogue I submit?
Results are cached on the server, addressed by the content itself, so re-submitting the same text doesn't re-run the pipeline. If you're signed in and save a result, it's linked to your account so you can find it again; if you're not signed in, nothing is tied to an identity.