Generates ElevenLabs text-to-speech audio with local playback via the sag CLI.
npx skills add https://github.com/shalevamin/Tau_agent --skill sag-shalevamin
Please help me install this Agent Skill. Skill: sag Source: https://github.com/shalevamin/Tau_agent/tree/main/tau-agent-main/skills/sag Command: npx skills add https://github.com/shalevamin/Tau_agent --skill sag-shalevamin
sag "Hello there", choosing voices with -v and listing options via sag voices. - Model & Delivery Control: Select models like eleven_v3, eleven_multilingual_v2, or eleven_flash_v2_5, and use v3 audio tags such as [whispers], [laughs], and [short pause] for expressive delivery. - Voice Chat Replies: Generate character-voice audio files (e.g., sag -v Clawd -o /tmp/voice-reply.mp3 "message") and attach them to chat responses. - Use Case: A user asks the agent to reply in a "crazy scientist voice"; the agent generates an MP3 with excited audio tags and dramatic pauses, then sends the audio file in the conversation. ## Quick Start Ask the agent to say a message out loud using sag, optionally specifying a voice or character style.High-intent search queries and answers about installing and using this skill.
Use the sag CLI: run `sag "Hello there"` to speak text aloud, or `sag -o output.mp3 "text"` to save audio to a file. You need an ELEVENLABS_API_KEY environment variable set before running commands.
List available voices with `sag voices`, then select one using the `-v` flag, such as `sag speak -v "Roger" "Hello"`. You can also set a default via the ELEVENLABS_VOICE_ID or SAG_VOICE_ID environment variables.
The default `eleven_v3` is the most expressive and supports audio tags like `[whispers]` and `[laughs]`. Use `eleven_multilingual_v2` for stability across languages, or `eleven_flash_v2_5` when speed matters most.
Support depends on the model: v3 does not support SSML `<break>` and instead uses tags like `[pause]`, `[short pause]`, and `[long pause]`. Models v2 and v2.5 support `<break time="1.5s" />`, but `<phoneme>` is not exposed in sag.
First respell the word, add hyphens, or adjust casing (e.g., "key-note"). For numbers, units, and URLs, use `--normalize auto`, and use `--lang en|de|fr` to bias normalization toward a specific language.
trpc-groupGenerates ElevenLabs text-to-speech audio with mac-style say command UX.
buda-aiExtract and enhance audio from video files for MP3/WAV/AAC output.
lvhuanidTranscribe spoken audio into written text using the z-ai-web-dev-sdk.
taiyousan15Convert Japanese text to speech using multiple TTS engines and output formats.