Adding Subtitles to Libro.fm Audiobooks with Whisper.cpp

March 5, 2026

Transcribe DRM-free audiobooks locally and read along

TLDR; Convert the audiobook to WAV, run whisper-cli to generate an SRT, then mux it back into the m4b with ffmpeg.

Libro.fm is a great place to buy audiobooks — it’s DRM-free, which means you get your audiobook as a standard m4b file that you actually own. I host mine on a home server and listen on any device I want.

The one thing Libro.fm audiobooks don’t come with is subtitles. I find it much more immersive to read along with the narration, so here’s how I add them using whisper.cpp.

Setup

Install whisper.cpp and a model. On Arch Linux via the AUR:

# With NVIDIA GPU (recommended)
paru -S whisper.cpp-cuda

# CPU-only fallback
paru -S whisper.cpp

For the model, whisper.cpp-model-small works well for most content. If you have more than 5.5 GB of VRAM, whisper.cpp-model-large-v3-turbo gives noticeably better accuracy.

Procedure

1. Convert the audiobook to a format whisper.cpp expects (16 kHz mono WAV):

ffmpeg -i input.m4b -ar 16000 -ac 1 -c:a pcm_s16le 16khz_audio.wav

2. Transcribe and generate an SRT subtitle file:

whisper-cli 16khz_audio.wav -l en -pp -ps --output-srt \
  -m /usr/share/whisper.cpp-model-large-v3-turbo/ggml-large-v3-turbo.bin

This produces 16khz_audio.wav.srt.

3. Mux the subtitle back into the original audiobook, preserving chapters and metadata:

ffmpeg -i input.m4b -i 16khz_audio.wav.srt \
  -map 0:a -map 0:v? -map 1:s \
  -map_metadata 0 -map_chapters 0 \
  -c copy -c:s mov_text \
  output.m4b

The resulting output.m4b has the embedded subtitle track and can be played back in any player that supports m4b subtitles, such as Prologue or VLC.