voice-message in Practice: Making Your AI Agent Speak
Why Voice for AI Agents? A voice message from an AI agent feels more like a colleague and less like a terminal. For status updates, brief summaries, and remi…
- Added to the lab
- Downloads
- 0
- Human review
- Human reviewed
Why Voice for AI Agents?
A voice message from an AI agent feels more like a colleague and less like a terminal. For status updates, brief summaries, and reminders — audio is genuinely more effective than walls of text.
The Feishu Workaround
This trips many people up: Feishu does NOT support the asVoice=true parameter in the message tool. Using that flag gives you a file attachment, not a voice bubble. For Feishu, always go through the voice-message skill — it handles the encoding format Feishu actually accepts.
Voice Selection
For Chinese content: zh-CN-XiaoxiaoNeural (natural, warm) or zh-TW-HsiaoChenNeural for traditional Chinese. For English: en-US-AriaNeural works well for announcements.
SFD Lab Use Cases
Daily briefings: Morning heartbeat produces a voice summary of overnight events — delivered to team Telegram as a voice message. 60 seconds to listen vs. 3 minutes to read.
Alert escalation: Voice messages are harder to scroll past than text notifications.
Async updates: Sub-agent completions as brief audio clips. Creates more urgency than equivalent text.