---
name: grok-tts
description: Discover and invoke the GROK_TTS directory capability using its live contract. Use when a request maps to this exact capability.
---

# GROK_TTS

This Skill is the behavioral expression of [the GROK_TTS directory article](/a/directory/GROK_TTS). The directory row remains the executable contract.

## Operate

1. Read GET /api/directory/GROK_TTS immediately before use.
2. Confirm the request is action, not explanation, and that authority permits the exact operation.
3. Follow the row's current ARGS and examples exactly; do not infer undocumented parameters.
4. Invoke POST /api/dispatch with {"key":"GROK_TTS","body":"<row-shaped args>"}.
5. Return real output and receipt; on failure, compare the row contract with the invocation before changing code.

## Human explanation

# WHAT: Text-to-SPEECH (Grok). INVOKE: [GROK_TTS]<text>|<voice_id>[/GROK_TTS] # WHEN_TO_USE: you need to grok tts # ARGS: see content # EX: [GROK_TTS]arg1|arg2|arg15[/GROK_TTS] # $1 = the words to speak; $2 = voice_id (eve|ara|rex|sal|leo). Returns mp3 audio BYTES (binary, not JSON) from https://api.x.ai/v1/tts; ~$15 per 1M characters. # To DELIVER spoken audio into a chat in ONE call, use [VOICE_SEND] instead (it synthesizes + attaches). {"text":"$1","voice_id":"$2","language":"en"}