Papermorph is a Claude skill that turns a PDF textbook into an animated, narrated, interactive web book

Turning a textbook that already exists as a PDF into something a learner reads on a phone, listens to and clicks through has usually been a studio job. Papermorph is a Claude Code skill that produces the same output from the PDF, and its author, who publishes under the handle DozenTwelve, released it as MIT-licensed open source on 3 October 2026.

The pipeline is spelled out in the project's README: PDF, to book plan, to storyboards, to narration, to animation and quizzes, to a web book. The output is a static site with one chapter per screen, a voiced read-through, animated figures and multiple-choice checks, and a bookshelf that groups several books into one page. The GitHub REST API reports 113 stars one day after the first public commit.

What is live and what is not

Two books are already on the public bookshelf at papermorph.diamonddoge.org: an elementary algebra book with 68 narrated chapters and an elementary mathematics book with 15. Both are open to any visitor and play in a browser with no sign-in. The README states that the current version requires the model Claude Opus 5.5 and that image generation, languages other than English and background music are not yet supported. Picture books and social-science titles are named as planned, not shipped.

The author's recommended first run asks the agent to produce one English chapter from a PDF for review before the rest of a book is rendered, which keeps the model's output in sight before a long run.

Why the skill matters for a working team

The project replaces three jobs that educational publishers usually hire out: writing a storyboard from a text, recording a narration track and building the interactive layer that holds all three together. Each of those costs money per minute of finished content, and they are usually scheduled against each other, so a late change on one resets the other two. A skill that reads a PDF and produces all three in one pass turns the unit of work from a chapter in a studio into a chapter the agent writes and an editor accepts or rejects on screen.

The output is a plain static site, which is a format a school or a company already knows how to host. The reference bookshelf runs out of a public web server with no build system in front of it.

The limits worth naming before adopting it

Papermorph is new. The GitHub REST API reports the first commit on 3 October 2026 and a star count of 113 one day later: a real signal that the project is being noticed, and not evidence yet that it survives the second textbook a user tries. The two public examples are both mathematics, where a diagram follows a predictable grammar; a humanities text with photographs and quotations is the harder case the author names as planned work.

The skill also depends on a single model today, so a team that wants to run Papermorph against a long book is tied to the pricing and availability of Opus 5.5 for as long as a chapter takes. A team that has a backlog of textbooks to convert should price one full book at current token rates before scheduling the rest.

Source

Primary source: DozenTwelve/Papermorph on GitHub. Live bookshelf: papermorph.diamonddoge.org.