| name | aima-tts |
| description | Text-to-speech using AIMA's current local TTS model. Generate audio files from text. |
| metadata | {"openclaw":{"emoji":"🔊","requires":{"bins":["curl"]},"always":true}} |
AIMA Text-to-Speech
Generate speech audio from text using the TTS model currently managed by AIMA/OpenClaw.
When to use
- The user explicitly asks for a spoken or voice reply.
- The user asks you to say something aloud instead of only writing it.
- The user wants a short audio clip in Chinese or English.
Required behavior
- Write a short spoken script first.
- Run
{baseDir}/scripts/speak.sh with that script.
- After OpenClaw attaches the generated media, reply with exactly
NO_REPLY.
- If TTS generation fails, give a brief text fallback that says the audio conversion failed.
- Keep the spoken script short and in the user's language unless they ask for something longer.
Quick start
{baseDir}/scripts/speak.sh "你好世界" --filename hello.wav
Useful flags
{baseDir}/scripts/speak.sh "今天天气真好" --filename weather.wav
{baseDir}/scripts/speak.sh "Hello AIMA" --filename greeting.wav --voice default
{baseDir}/scripts/speak.sh "请用参考音色说这句话" --api tts --reference-audio /tmp/ref.wav --reference-text "参考文案"
Output
- WAV audio file saved to workspace
MEDIA: line printed for OpenClaw auto-attachment
Notes
- Model: auto-detected from
~/.openclaw/openclaw.json (override with AIMA_TTS_MODEL)
- Voice:
default (single voice)
- API:
speech (/v1/audio/speech) or tts (/v1/tts)
- Output format: configurable with
--response-format
- Optional reference fields:
--reference-audio, --reference-text
- Runs on AIMA proxy at
http://127.0.0.1:6188/v1