EdgeSpeak

Agent Skill

Let Skills-capable agents discover local speech workflows.

Install

Install with the standard skills installer — let it detect the client, or pass an explicit agent target.

npx skills add lattifai/EdgeSpeak

npx skills add lattifai/EdgeSpeak --agent claude-code
npx skills add lattifai/EdgeSpeak --agent cursor
npx skills add lattifai/EdgeSpeak --agent codex

Skills

Eight skills covering the on-device pipeline end to end. Most read the skill instructions and shell out to edgespeak-cli, returning JSON in the same shape as the gateway API. Karaoke also needs FFmpeg, yt-download needs yt-dlp, and translate is written by the agent itself with no CLI behind it.

edgespeak-transcribe      audio/video → text + word timing   (edgespeak-cli transcribe)
edgespeak-align           known text → word timing           (edgespeak-cli align)
edgespeak-segment         raw text → sentences               (edgespeak-cli segment)
edgespeak-broadcast       text → speech + voice library      (edgespeak-cli speech)
edgespeak-karaoke         media → karaoke ASS + hard subs    (edgespeak-cli transcribe + FFmpeg)
edgespeak-name-speakers   speaker_N → real names             (bundled Python script)
edgespeak-translate       timed transcript → new language    (agent-written, no CLI)
edgespeak-yt-download     YouTube URL → media + captions     (yt-dlp)

Requirements

The skills shell out to edgespeak-cli — a self-contained on-device runtime, no desktop app required. The desktop app ships the same CLI.

curl -fsSL https://edgespeak.com/install.sh | sh
edgespeak-cli trial

Manual install

If the installer is not available, clone the public repository and copy the skill folders into your agent's skills directory.

git clone https://github.com/lattifai/EdgeSpeak
cp -r EdgeSpeak/skills/edgespeak-* ~/.claude/skills/