Skip to content

Repository files navigation

YOOS-APP — Bring Your Own Voice

CI Python 3.10+ License: MIT

Feed it your own writing. It learns your voice. Then it writes in that voice — consistently, at scale.

Most "AI writing" tools sound like everyone else's AI writing: flat, generic, instantly recognizable. YOOS-APP solves the opposite problem. You give it a folder of texts you have written; it builds a portable profile of how you write — your sentence rhythm, your punctuation habits, your vocabulary, your openings — and then generates new content that stays in your voice.

It's built for people who have something to publish but find writing slow or hard: solo founders, bloggers, small brands, newsletter authors, anyone who needs a steady stream of on-brand content without sounding like a robot.


Why it exists

Writing consistently is the bottleneck. Not everyone can sit down and produce clean, on-voice copy every week — but everyone has a backlog of things worth saying. Generic LLM output doesn't help: it's correct but voiceless, and readers feel it.

YOOS-APP turns your existing writing into a reusable voice asset. Analyze once, generate forever.

  • No embeddings. No vector database. No cloud lock-in. Voice is captured as plain, inspectable statistics in a portable JSON file.
  • Runs fully local with Ollama, or with any major API (OpenAI, Anthropic).
  • Auditable output — every generation gets a 0–100 voice-match score so you know how close it landed.

What it does

   Your corpus (.txt / .pdf / .html)
            │
            ▼
     ┌─────────────┐    25+ stylometric dimensions:
     │   analyze   │    sentence-length distribution, first-person rate,
     │             │    punctuation fingerprint, vocabulary richness,
     └──────┬──────┘    paragraph rhythm, signature phrases…
            │  voice profile  →  portable JSON
            ▼
     ┌─────────────┐    Genre templates: blog, guide, magazine,
     │  generate   │    news, story, column, essay, marketing
     │  + backend  │    Backends: OpenAI · Anthropic · Ollama (local)
     └──────┬──────┘
            │  draft + voice-match score (0–100)
            ▼
     ┌─────────────┐
     │   export    │    → file · PDF · WordPress
     └─────────────┘

Quickstart

git clone https://github.com/yoldaolmak/YOOS-APP.git
cd YOOS-APP
pip install -r requirements.txt

# See it work end-to-end, no setup:
python -m yoos_app demo

Use it on your own writing

# 1. Learn your voice from a folder of your texts
yoos-app analyze --corpus ./my-articles/ --author "Me" --out my-voice.json

# 2. Generate a new piece in that voice
yoos-app generate --profile my-voice.json --type magazine \
                  --topic "Three days in Lisbon" --backend openai

# 3. Or run the whole pipeline at once and publish
yoos-app run --corpus ./my-articles/ --type travel_blog \
             --topic "Why I keep going back to Tokyo" --dest wordpress

That's the whole loop: analyze → generate → export.


Who it's for / what it unlocks

You are… YOOS-APP gives you…
A solo founder / indie brand On-voice blog & marketing copy without hiring a writer
A blogger or newsletter author A way to keep publishing on the weeks you can't write
An agency One reusable voice profile per client, consistent across drafts
A non-native or reluctant writer Your ideas, in clean prose that still sounds like you

The opportunity is consistency at scale: capture a voice once, then produce dozens of pieces that read like the same human wrote them.


How the voice profile works

The analyzer is pure, transparent text statistics — no black-box embeddings. It measures things like:

  • Sentence-length distribution (mean, spread, short/long ratio)
  • First-person, question, and exclamation rates
  • Paragraph rhythm (sentences per paragraph)
  • Punctuation style (em-dash, semicolon, parenthesis use)
  • Vocabulary richness (type–token ratio)
  • Signature 2-gram phrases and typical openings

The result is a small JSON file you can read, diff, version, and reuse anywhere.


Backends

Backend Use it when
Ollama (local) You want zero API cost and full privacy
OpenAI You want top-tier quality out of the box
Anthropic You prefer Claude models

Set your key in .env (see .env.example). No key needed for the local Ollama path or the demo.


Project status & sibling project

YOOS-APP is the accessible, bring-your-own-voice entry point: give it your texts, get consistent content in your voice.

graphova is its more advanced sibling — a voice-print engine for writing in any author's tone. YOOS-APP is the simple front door; graphova is the deep end.


License

MIT — see LICENSE. Use it, fork it, build on it.

About

Universal Author Voice Engine — upload any author's texts, extract their voice, generate content in that style. PDF/HTML/TXT in & out. OpenAI, Claude, Ollama, Codex.

Topics

Resources

Contributing

Security policy

Stars

3 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages