Why a Reader App?

Obsidian Metadata

categories

Project

Inspiration:

Andrej Karpathy wrote a post about how reading within EPUB could be more effective with the use of LLMs. He had pulled an MVP out, which was just a basic code he wrote in Python, and I wanted to build on this idea. Why not have all of what he is talking about and more happen within the app itself?

Tweet : Github repo : https://github.com/manghat/reader3

Other products

• Readever - Product Hunt
  • Readever
    • What I like
      • Hooks
        • reading in 2025 shouldn’t be alone
        • Finish more books
        • you can actually interact with them
        • Passive reading into active understanding
      • Videos
      • Can chat with celebs around books
      • Discover new books
      • Probably going to be #1 on Product Hunt

Personal Need

I’ve been trying to get back into reading, but I’ve found it hard to actually read physical books or even read on a screen. I find I stick to my reading much more when I can switch between listening and reading. This is especially true for books with graphics, as it makes more sense to have the same screen with a text-to-speech module that can read it aloud when needed.

Version 1

Link

Unbound - Read, But also listen to epubs for free

Research

• Upload an epub and a m4b audiobook and create a space where the audiobook and the book can be read together.
  • Resources :
    • https://readium.org/mobile/
    • https://github.com/kanjieater/SubPlz?tab=readme-ov-file
      🫴 Generate, sync, and manage subtitle files for any media; Generate your own audiobook subs similar to Kindle’s Immersion Reading 📖🎧
    • https://news.ycombinator.com/item?id=38747710
• Opinions
  • Reddit post
    • Oh, look, yet another sketchy group in the TTS space trying to get you to download a bunch of pickled files, that can hide malicious code.
    • It’s almost like a right of passage for TTS models: see how many suckers on LocalLLaMA you can get to download your pickled files.
  • epub to audiobook
    • Github - https://github.com/p0n1/epub_to_audiobook
• TTS
  • https://www.hume.ai/
• Roadmap Features
  • TTS not working on iOS.
    • Provide a server side tts option.
  • Recap till your point
    • When I pick up a book I left off earlier, most frustrating part is to remember what happened.
  • Thoughts
    • Highlights are useless for me. What if we have thoughts that can be captured on the book itself?
• Build
  • Voice Generation options
    • Review of available options
    • There is a free option to use browser capability
      • SpeechSynthesisUtterance
        • https://developer.mozilla.org/en-US/docs/Web/API/SpeechSynthesisUtterance
        • Video explainer - https://www.youtube.com/watch?v=nx_k1XCaWWs
      • ‘Listen to Page’ chrome feature - deepdive
    • Models
      • Compare all the options—Demo—blog
      • Style - demo - github
        • Sounds good.
      • ChatterBox - demo - server - api+web
        • Sounds very good.
      • Kokoro - web - code - Demo
        • Uses Kokorojs
  • supabase bucket policies
    • Read - select
      • — Allow anyone to read book assets (since they’re served through our API)
      • CREATE POLICY “Allow public read access to book assets” ON storage.objects
      • FOR SELECT USING (bucket_id = ‘book-assets’);
  • Sentence Splitting solutions found:
    • sentence-splitter - package - with tests - link
    • langchain - https://python.langchain.com/docs/how_to/split_html/
    • regex for selecting html elements in text - link
      • /<\/?\w+((\s+\w+(\s*=\s*(?:".*?"|'.*?'|[\^'">\s]+))?)+\s*|\s*)\/?>/
  • Debugging