fliphtml5-liberator

fliphtml5-liberator

Command-line extractor for FlipHTML5 books. Resolves the page manifest, downloads page images at the highest available resolution, and assembles them into a single PDF.

FlipHTML5 publishes books as JavaScript-driven flipbook viewers backed by a config.js file that lists the pages. Recent / "Protected" deployments obfuscate that page list with an Emscripten-compiled WebAssembly binary (deString.wasm) loaded by deString.js. This tool handles both layouts.

Capabilities

Capability	Status
Plain config.js (page list exposed in source)	✅ Direct extraction, no Node.js needed
Encrypted / "Protected" config.js (WASM-decoded)	✅ Via host-environment polyfill in Node.js
Nested encryption (page list itself an encrypted blob)	✅ Recursive WASM decode
Hashed image filenames (/files/large/<hash>.webp)	✅ Resolved from manifest
Concurrent page downloads	✅ httpx.AsyncClient + asyncio.gather
WebP and JPEG inputs	✅ Auto-detected; WebP converted via Pillow before PDF assembly
PDF output	✅ Single book.pdf via img2pdf

How extraction works

The downloader runs in two stages.

Stage 1: Manifest extraction

Fetches https://online.fliphtml5.com/<book_id>/javascript/config.js (falling back to /config.js) and tries the plain-text path first:

1. Plain path. Regex search for the fliphtml5_pages = [...] assignment. If present, the JavaScript object literal is normalized to JSON and parsed directly. No subprocess required.
2. WASM path. If the page list is absent or stored as an encrypted string, the script spawns node fliphtml5_decoder.js <config_path> and reads the decrypted manifest from stdout.

The Node.js helper performs the decryption inside a host-environment polyfill so the original WASM binary runs unmodified:

• Loads deString.wasm as a binary buffer.
• Patches deString.js to remove the inlined data URI loader and replaces the Module.onRuntimeInitialized callback so initialization resolves a Promise.
• Reimplements stringToUTF8 and UTF8ToString against the WASM heap because the upstream exports are typically stripped or mangled.
• Evaluates the original config.js to materialize whatever bookConfig / htmlConfig globals it defines.
• Calls the exported _DeString() to decrypt bookConfig and the fliphtml5_pages string.
• Detects nested encryption (decoded payload starting with the v signature) and re-runs _DeString() until the result parses as JSON.

The decrypted manifest is written to stdout and consumed by Python.

Stage 2: Page download and PDF assembly

For each page entry in the manifest, the script resolves an image URL using the entry's l (link) or n (number) field:

• Absolute URLs are used as-is.
• Paths starting with files/ are prefixed with http://online.fliphtml5.com/<book_id>/.
• Bare hash-style filenames are placed under http://online.fliphtml5.com/<book_id>/files/large/<filename>, which is the highest-resolution variant FlipHTML5 publishes.

Pages are downloaded concurrently with a 15-second per-request timeout via httpx.AsyncClient. Failed pages are logged as warnings and skipped (the resulting PDF will be missing those pages rather than aborting the entire run). WebP responses are converted to PNG with Pillow before img2pdf.convert() writes book.pdf.

Prerequisites

• Python 3.7+
• Node.js (any recent LTS) for the WASM decoder path. Not required if every book you process exposes its page list in plain config.js.
• deString.js and deString.wasm from the target book, placed in the project root. These are not bundled because they are the book publisher's binaries and may vary between FlipHTML5 deployments.

Acquiring deString.js and deString.wasm

1. Open the target book in a browser.
2. Open DevTools (F12) and switch to the Network tab.
3. Reload the page.
4. Filter requests for deString.
5. Right-click each of deString.js and deString.wasm and save them into the fliphtml5-liberator/ directory next to downloader.py.

The same pair generally works across books published by the same FlipHTML5 account version. If decoding fails on a new book, refresh the binaries from that book's network trace.

Installation

git clone https://github.com/silenthooligan/code-sharing.git
cd code-sharing/fliphtml5-liberator

pip install httpx img2pdf Pillow

There is no requirements.txt; the dependency surface is small and stable. Pin manually if reproducibility matters to you.

Usage

python downloader.py <book_url_or_id>

Either a full URL or just the <account>/<book> ID portion is accepted:

python downloader.py https://online.fliphtml5.com/ousx/stby
python downloader.py ousx/stby

Output goes to book.pdf in the current working directory. Intermediate image files are written to a temp directory that is cleaned up on exit. Logs go to stdout in LEVELNAME: message format.

Operational notes

• Output filename is hardcoded. Each run overwrites book.pdf in the current directory. Run from a per-book working directory or rename after each invocation.
• Non-ASCII config.js. Some books ship config.js with non-ASCII bytes (often book metadata in CJK or accented Latin). The downloader reads with errors='replace' so this no longer aborts with UnicodeDecodeError; the regex matchers operate on the replaced text and find the page-list assignment unchanged.
• WASM path is skipped automatically when the plain page list is present, avoiding subprocess overhead for the common case.
• Hashed filenames may 404 if the book uses a non-standard storage layout. The warning is logged and the page is skipped; inspect the manifest entry to see whether the l field contains a usable fallback path.
• Concurrency is unbounded. asyncio.gather issues all page requests in parallel. For large books on slow links, consider patching download_image to use a Semaphore.

Disclaimer

For educational and personal-archival use only. Respect the copyright and terms of service of any content you process. Do not redistribute copyrighted material without permission from the rights holder.

License

MIT.