voice-chat
Skill0 starsChanged 3 months agoMITView on GitHub
Voice conversation with the user via push-to-talk. Use whenever the user asks to "turn on voice mode", "bring up voice chat", "let's talk", wants to speak instead of type, or otherwise wants a spoken back-and-forth. The assistant speaks via TTS and hears the user via push-to-talk Whisper.
- Reads credentials
2 other projects use a version of this.
Adopted by
Most recently taken up by
- jacquardlabs/voice-suitetheir own version, first seen Jul 11, 2026
- tomo3141592653/ayumu-osstheir own version, first seen Mar 7, 2026
Other versions2 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 2
- tomo3141592653/ayumu-oss★ 9same name
- jacquardlabs/voice-suite★ 3same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repositoryOne other file it gives its agents.
- setupSkill
The file
We split this skill into the rules it gives an agent
4 rules. About 0 are common, 1 is distinctive, and 3 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
3
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 57
Keep lines conversational and moderate length — this is speech, not a document.
- Distinctive: line 62
No markdown, bullets, or emoji in the say-file — it's read aloud verbatim.
- Very specific: line 73
If the tool errors about missing modules/endpoints, the plugin may not be set up.
- Very specific: line 74
Run scripts/install.sh from the repo root (in Claude Code: the setup skill, /push-to-talk:setup), and make sure …
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Push-to-talk voice chat
You talk with the user out loud. Each turn: write your line to a scratch file,
then run the bundled ptt tool to speak it and (optionally) capture the user's
spoken reply.
Running a turn
The tool is ptt — on PATH after scripts/install.sh. (In a Claude Code
plugin install it also lives at ${CLAUDE_PLUGIN_ROOT}/bin/ptt; use that
absolute path if plain ptt isn't found.) Three subcommands:
speak <sayfile>— speak the text in<sayfile>. No microphone. No special permissions.listen— push-to-talk capture → prints the transcript to stdout.chat <sayfile>—speakthenlisten(one full back-and-forth turn).
listen and chat read the keyboard directly (evdev), so they MUST run under
the input group:
sg input -c 'ptt chat /abs/path/to/say.txt'speak does not need sg:
ptt speak /abs/path/to/say.txtStep by step
- Write your spoken line to a scratch file using your file-write tool (this avoids
all shell-escaping of quotes/newlines). Use a temp path you control, e.g. the
session scratchpad, or
/tmp/ptt-say.txt. Pass that path as the<sayfile>argument. Do NOT try to inline the text on the command line. - To speak AND get a reply, run
chatundersg input. To just make a statement and keep working (no reply expected), runspeak(nosg). - Give the shell call a long timeout (e.g. 10 minutes) — the user may take a while to respond, and TTS has startup latency.
Push-to-talk, tell the user once
- Hold the PTT key (default Right Alt) and speak; release to send.
- Tell them to keep holding through pauses — releasing mid-thought cuts them off.
- They do NOT hold the key while you're talking, so there's no echo.
Style for spoken lines
- Keep lines conversational and moderate length — this is speech, not a
document. Split a long update into a few
speakcalls rather than one wall of text. TTS can have noticeable startup latency per line. - Spell out things TTS mangles (URLs, code) phonetically: "example dot com", "H T M L".
- No markdown, bullets, or emoji in the say-file — it's read aloud verbatim.
Exit codes (from the tool, on stderr)
3— no readable keyboard: the user isn't in theinputgroup. Tell them to runsudo usermod -aG input $USER(then relogin, or you re-run undersg).4— no key press within the timeout: the user didn't hold the key. Just try the turn again, or ask if they're there.
First-run
If the tool errors about missing modules/endpoints, the plugin may not be set up.
Run scripts/install.sh from the repo root (in Claude Code: the setup skill,
/push-to-talk:setup), and make sure ~/.config/claude-ptt/config.env has the
user's TTS/STT URLs.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

