<- all tokdocs

Papermorph: The Claude Skill That Turns PDFs Into Web Books

By , founder of Mavric and Apso

View on TikTok ->

Papermorph is a free, MIT-licensed skill for Claude Code that turns a PDF into an animated, narrated web book with in-browser quizzes. It is real and already has 88 finished chapters online. Two claims in the TikTok that spread it are off: the voice comes from Microsoft Edge text-to-speech rather than the AI model, and "no image models" is a gap the author plans to fill.

What is Papermorph?

The README describes it as "A Skill that turns PDFs into animated interactive web books." The pipeline it lists is "PDF -> Book plan -> Storyboards -> Narration -> Animation & quizzes."

The repository is DozenTwelve/Papermorph, created on October 3, 2026. On October 6 it had 385 stars, 42 forks and 1,621 files, 1,406 of them MP3 narration clips. The live bookshelf at papermorph.diamonddoge.org lists two finished titles: "Elementary Algebra" with 68 narrated chapters and "Elementary Mathematics" with 20. That makes 88 chapters of generated HTML and audio.

Each chapter is "one 1600×900 SVG stage plus controls." A lesson in the algebra book draws "3 + 5 = 8" and "5 + 3 = 8" as colored arcs over a number line to teach order in addition.

A TikTok from @github.signals introduced the project two days after the repository was created.

Does Claude generate the narration?

No. The narration comes from Microsoft's Edge text-to-speech service through the edge-tts Python package. The top of scripts/tts.py says: "Generate narration audio and word-mark timings with Edge TTS." The script records word boundary events so each animation beat can wait for a specific spoken word.

This follows from what Claude can output. Anthropic's models overview says "All current models support text and image input, text output, multilingual capabilities, vision, and tool use." Text output only means the voice has to come from somewhere else. Edge TTS is free and needs no API key, but it does need network access to a Microsoft endpoint, and it is the one part of the pipeline the author does not control.

SKILL.md lists the requirements: "Needs: uv, ffmpeg/ffprobe, Edge TTS network access and Playwright Chromium."

Is "no image models" a feature?

The video calls it "the most surprising part." The README treats it as a limitation:

Today: Opus 5.5 only. No image models, multilingual support, or BGM yet.

Planned: Image models and storyboarding for interactive picture books and humanities documentaries.

Missing image models sit in the same list as missing multilingual support and missing background music. All three are things the author intends to add. The current visuals are hand-coded SVG, and SKILL.md says "Default to code/SVG artwork; record any other asset choices in the book's conventions."

Which Claude model does it use, and what does it cost?

The README says "Opus 5.5 only." Anthropic's models overview lists Claude Opus 5.5 (claude-opus-5-5) as current and recommends it: "If you're unsure which model to use, start with Claude Opus 5.5 for most workloads."

The skill is free. The model costs $4 per million input tokens and $20 per million output tokens. Each shipped chapter is an HTML file of 22 KB to 27 KB, written one at a time, and SKILL.md recommends "one fresh subagent per chapter sequentially." A full book means many Opus calls, which the video does not mention.

Do the quizzes work?

Yes. The engine reference documents a quiz() call with choice, tap and blanks question builders, graded in the browser through api.grade(). The first algebra chapter ships a working quiz that asks the reader to tap "the whole number that is not a natural number" on a number line. There is no backend, and the bookshelf notes "Reading progress stays in your browser."

How to use Papermorph

The install command shown in the README is npx skills add DozenTwelve/Papermorph --skill papermorph. The suggested prompt begins: "/papermorph Turn /path/to/book.pdf into an animated interactive web book. Target readers: [your audience]. Start with one English chapter for review."

You need Claude Code with Opus 5.5, uv, ffmpeg and Playwright Chromium installed with uv run --with playwright playwright install chromium.

The process involves several approvals from you. SKILL.md's six stages include an intake step ("Ask for missing decisions together, with defaults"), a chapter map you confirm, and a pilot chapter delivered "for the user's review." For every chapter, "The user reviews overall aesthetics, pacing and teaching effectiveness." The video's "you just feed it a book" leaves this out.

The video shows the link as github.com/dozentwelve/papermorph. GitHub accepts that casing in a URL, but the install command needs DozenTwelve/Papermorph.

What is on the bookshelf

The demo bookshelf shows six titles, and only two exist. The site marks physics, the picture books ("Little Bear in the Woods", "Where Are the Stars?") and the curious books ("Why Is It So Fun?") as "Coming soon," and states "Books on the future shelves are previews, not available lessons."

The GitHub account behind the project uses a pseudonym, and I found no profile or company to attribute it to a named person.

Bottom line

Papermorph works and has shipped 88 narrated chapters with working quizzes. The voice is Microsoft Edge TTS, not Claude, and the lack of image models is a stated gap on the roadmap. The skill is free, but each book runs on Opus 5.5 at $4 and $20 per million input and output tokens, with a human reviewing each chapter.

Sources

Published October 5, 2026. Based on a TikTok by @7614639676569895957. Full video notes (markdown)