skills.rest
agentprojectcontext
agentprojectcontextOFFICIAL
Explore all skills by this author →
WebsiteGitHub

apx-voice

Configure text-to-speech engines and voice pipelines for APX agent spoken output.

6|1|Updated May 8, 2026
Search Tags:#voice#text-to-speech#audio#gemini#elevenlabs#tts#piper
One-click install
npx skills add https://github.com/agentprojectcontext/apx --skill apx-voice-agentprojectcontext
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: apx-voice
Source: https://github.com/agentprojectcontext/apx/tree/main/src/core/runtime-skills/apx-voice
Command: npx skills add https://github.com/agentprojectcontext/apx --skill apx-voice-agentprojectcontext

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? APX agents reply in text by default, and getting them to speak requires choosing a TTS engine, wiring API keys, handling audio formats, and debugging silent output. This Skill provides the operational knowledge to configure, test, and troubleshoot the full APX voice stack. ## Core Features & Use Cases - Multi-engine TTS configuration: Set up Piper (local/offline), ElevenLabs, OpenAI, Gemini, or custom OpenAI-compatible endpoints like QVox/Qwen3-TTS via voice.tts.* config keys. - Emotion tags and voice channel: Enable per-engine [tag] emotion markers, and use the unified POST /api/voice/turn endpoint for mic-to-speaker round trips combining Whisper STT, the agent loop, and TTS. - Use Case: You want APX to read Telegram replies aloud while driving. Configure the Gemini engine with an API key, enable mobility replies, and every automatic reply goes out as an OGG/Opus voice note followed by a text transcript. ## Quick Start Ask the agent to configure Piper as the local TTS engine and run a test phrase with apx voice say to confirm spoken output works.

Frequently Asked Questions about apx-voice

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I make APX speak replies out loud?▼

Set a TTS provider with apx config set --global voice.tts.provider, supply the engine's API key or model path, then test with apx voice say. The --global flag is required because voice config is only read from the global config file.

Which TTS engine should I use for offline text-to-speech?▼

Piper is the local offline option requiring no API key, just the piper binary and an .onnx voice model such as es_AR-daniela-high. Cloud alternatives are ElevenLabs, OpenAI tts-1, and Gemini, each needing an API key.

Can I use a custom OpenAI-compatible TTS server with APX?▼

Yes, define voice.tts.custom.<slug> with a base_url pointing at any OpenAI-compatible /v1/audio/speech server such as QVox or Qwen3-TTS. It uses only its own api_key and never leaks the OpenAI key.

Why is APX voice output silent or not playing?▼

Run apx voice providers to check availability, then apx voice say with --no-play and inspect the file with the file command. Gemini returns raw PCM wrapped in WAV, and playback needs a system player like afplay, paplay, or ffplay.

Does Gemini TTS return MP3 audio files?▼

No, Gemini TTS returns raw L16 PCM which APX wraps into a WAV container, producing .wav files with audio/wav mime type. Convert with ffmpeg if you need MP3.

How do emotion tags work in APX voice output?▼

Engines like QVox and Gemini support inline [tag] markers that switch speaking emotion per segment when voice.tts.<engine>.emotions.enabled is true. On engines without tag support, stray tags are stripped and never read aloud.

Related Skills

Explore More Software Engineering Skills→
zhaosj0315zhaosj0315

add-voice-transcription

Transcribes WhatsApp voice notes into text using OpenAI Whisper in NanoClaw.

Community
Advanced
bubbuildbubbuild
1.6k

telegram

Send, reply to, and edit Telegram messages via the Bot API.

Official
Intermediate
framerslabframerslab

streaming-stt-whisper

Transcribe live speech in real time using the OpenAI Whisper API.

Official
Intermediate
AIDotNetAIDotNet
84

text-to-speech

Convert input text into speech audio files using pyttsx3.

Official
Basic