abogen

// Overview

abogen is a desktop text-to-speech tool that turns ePub, PDF, plain text, markdown and subtitle files into narrated audio with matching subtitles. It runs the Kokoro-82M model locally, so a whole book is converted on your own machine without a per-character API bill or an upload step. Output can be a single file or one file per chapter, in WAV, FLAC, MP3, Opus, or M4B with chapter markers.

The interesting part for most users is the caption side: abogen emits subtitles alongside the audio at a granularity you choose — per line, per sentence, per sentence-plus-comma, or in fixed groups of one, two, three or more words — in SRT or several ASS layouts. That makes it as useful for captioned short-form video voiceover as it is for long-form audiobooks.

Two interfaces ship in the same package and they are not identical. `abogen` opens a PyQt6 desktop GUI with the stable core feature set; `abogen-web` starts a Flask web UI that carries newer work, including Supertonic TTS, LLM-based text normalisation and Audiobookshelf integration, which the project is still folding back into the desktop app. abogen is MIT-licensed, published on PyPI, and runs on Windows, macOS and Linux with optional CUDA or ROCm acceleration. It depends on espeak-ng, which is installed separately.

// What it does

• Reads ePub, PDF, TXT, Markdown, SRT, ASS and VTT input, or text typed into the built-in editor
• Local synthesis with Kokoro-82M — no hosted API call and no per-character cost
• Synchronised subtitles at a granularity you pick: line, sentence, sentence + comma, or fixed 1/2/3-word groups, in SRT or four ASS layouts
• Chapter handling for books: choose chapters (or chapters + pages for PDFs), save each as its own file, merge them, or emit M4B with chapter markers
• Voice mixer for blending voice models into a saved custom profile, with preview before you commit to a long run, and speech rate from 0.1x to 2.0x
• Queue mode for batch conversion with per-file settings; a web UI (`abogen-web`) adds Supertonic TTS, LLM normalisation and Audiobookshelf integration

// Getting started

abogen installs from PyPI. espeak-ng is a separate prerequisite on every platform, and the GPU variant you install depends on your hardware.

# macOS
brew install espeak-ng

# Linux
sudo apt install espeak-ng    # Ubuntu/Debian
sudo pacman -S espeak-ng      # Arch
sudo dnf install espeak-ng    # Fedora

# NVIDIA on Windows (CUDA 12.8)
uv tool install --python 3.12 abogen[cuda] \
  --extra-index-url https://download.pytorch.org/whl/cu128 \
  --index-strategy unsafe-best-match

# CPU, or NVIDIA on Linux
uv tool install --python 3.12 abogen

# AMD on Linux (ROCm 6.4)
uv tool install --python 3.12 abogen[rocm] \
  --extra-index-url https://download.pytorch.org/whl/nightly/rocm6.4 \
  --index-strategy unsafe-best-match

abogen
# or
abogen-web

uv tool install --python 3.13 abogen \
  --with "kokoro @ git+https://github.com/hexgrad/kokoro.git,numpy<2"

Commands and code are distilled from the project's own documentation — always check the official repo for the latest.

// When to use it

• Convert an ePub or PDF you own into an M4B audiobook with chapter markers, entirely offline
• Generate voiceover plus word-level captions for short-form video without paying per character to a hosted TTS API
• Batch-narrate a folder of documents through queue mode, with different voices and settings per file
• Build a custom narrator voice in the voice mixer and reuse the saved profile across a whole series

// How abogen compares

abogen alongside other open-source audio, music & voice tools AI/TLDR tracks, ranked by GitHub stars.