See it in Claude Code
Recorded in a real session. The voice reads the highlighted sentence on screen.
/read-along.The whole walkthrough runs 1 minute 25.
Who it's for
Anyone who would rather hear long replies than read them: people with dyslexia, low vision, ADHD or eye strain, people reading in a second language, and anyone who wants to look away from the screen while Claude explains something. It is not a screen reader and doesn't replace one.
What it does
- Play any message. Hover a message for its play, speed and voice buttons.
- Follow the highlight. The sentence being spoken is highlighted in the message.
- A player above the prompt. Pause, stop, skip back or forward, and change speed and voice.
- Auto-read. Have each reply read when it arrives, or every update Claude writes between tool calls too.
- Several languages in one reply. Japanese, Chinese, Korean, Hebrew, Arabic, Cyrillic, Hindi, Greek and Thai sentences each get their own voice.
- Local and private. Speech is made on your machine. Nothing you read is sent anywhere.
- Fast without chipmunk voices. Speeds up to 4× keep the voice's pitch.
Languages
Read-along uses whatever voices your machine has, and picks one for each sentence, so a reply that mixes languages is read in each of them.
- Kokoro, the natural voices: English (US and UK), Japanese, Chinese, Spanish, French, Hindi, Italian and Portuguese. See Kokoro's voices.
- The macOS voices: dozens of languages, with Enhanced and Premium voices as free downloads. See Apple's guide.
- The Windows voices, from WSL: the languages added under Settings → Time & language → Speech.
- eSpeak NG: about 100 languages, robotic but on any Linux. See its languages.
Korean, Hebrew, Arabic, Russian, Greek, Thai and the rest are read with the system voices or eSpeak NG.
Hear it
More Kokoro voices, made with read-along's own setup on a laptop CPU, in other languages too. The sentence is highlighted while it plays.
-
English (US), Heart
kokoro:af_heartRead-along reads Claude's replies aloud, and highlights each sentence as it goes, so you can follow with your eyes and ears together.
At 2× speed, same pitch: -
English (US), Fenrir
kokoro:am_fenrirRead-along reads Claude's replies aloud and highlights each sentence as it goes, so you can follow with your eyes and ears together.
-
English (UK), Emma
kokoro:bf_emmaRead-along reads Claude's replies aloud, and highlights each sentence as it goes, so you can follow with your eyes and ears together.
-
Japanese, Alpha
kokoro:jf_alpha読み上げ中の文がハイライトされるので、目と耳で一緒に追うことができます。
-
Spanish, Dora
kokoro:ef_doraEste complemento lee en voz alta las respuestas de Claude y resalta cada frase mientras la lee.
-
French, Siwis
kokoro:ff_siwisCe module lit à voix haute les réponses de Claude et surligne chaque phrase pendant la lecture.
What you need
- Claude Code 2.1.287 or later, in the terminal. Mods also draw in the JetBrains plugin and the desktop app's Code tab, but not in its WSL sessions or over Remote Control.
- WSL, Linux or macOS. Native Windows isn't supported yet; use it through WSL. So far it's been used day to day on WSL; on Linux and macOS it passes its automated tests but hasn't been tried by hand yet, so reports are welcome.
- Python 3.9 or newer as
python3, and an audio player:ffplay(comes with ffmpeg) or SoX'splay. - A voice engine. Kokoro sounds the most natural, and read-along can install it for you (about 1 GB; a CPU is fine). The macOS and Windows voices and eSpeak NG work too.
Install
At the Claude Code prompt:
/plugin install read-along --marketplace tomfluff/claude-read-along
Answer y to add the marketplace, then pick the user scope. Then see what your machine has, and set up the Kokoro voices with Claude's help:
/read-along doctor
/read-along setup
Setup asks before it installs anything, and puts Kokoro in its own folder and nowhere else.
Using it
Hover a message and press play, or turn on auto-read with /read-along auto replies all. The player above the prompt pauses, stops and skips; with it focused (ctrl+x tab), p pauses, s stops, b and n go back and forward. /read-along opens the settings: a voice per language, speed, volume and auto-read.
The README has every command and setting, and what to do when something doesn't work.