飞书语音回复

Generate Feishu-native voice replies with a playable pause/resume bar by synthesizing text, converting it with ffmpeg to Ogg/Opus, and sending it as a voice...

Feishu Voice Reply

Create Feishu replies that contain both normal text and a real voice bubble/player. Avoid sending raw mp3 files when the goal is in-chat playback.

Workflow

1. Draft the text reply first.
2. Keep the visible text and the spoken text aligned.
3. Use scripts/build_feishu_voice.py to synthesize Chinese speech with Edge TTS and convert it to Ogg/Opus through ffmpeg.
4. Send the text reply normally.
5. Send the generated .ogg file with the messaging tool as a voice message:
   • mimeType: audio/ogg
   • asVoice: true
6. If a message id is available, reply to the triggering message for both the text and the voice.

Defaults

• Default language: Chinese
• Default voice: zh-CN-XiaoxiaoNeural
• Default channel behavior: Feishu text message + Feishu voice message

Rules

• 语音文件必须存放在 Workspace 下：使用 /root/.openclaw/workspace/temp/voice/ 目录，不要用 /tmp。否则飞书只显示文件路径而不是语音播放条。 <<<<<<< HEAD =======
• Emoji 自动过滤：脚本会自动过滤 emoji 字符再合成 TTS，显示的文本保留 emoji，但语音不会朗读 emoji（避免杂音）。无需手动处理。

8d2abf78b8490403831aae82052e8e107054b856

• Always show the text reply as well; do not send voice-only unless the user explicitly asks for that.
• Prefer concise spoken text. If the full answer is long, speak a compact summary and keep the full text visible.
• Use this skill specifically for Feishu playback UX. If the user only wants downloadable audio, normal attachments are fine.
• If synthesis succeeds but voice sending fails, tell the user clearly that Feishu voice delivery failed instead of pretending it worked.

Script

Build a Feishu-compatible voice file

Run:

python3 scripts/build_feishu_voice.py \
  --text "今天上海多云，气温十一度。" \
  --out-dir /root/.openclaw/workspace/temp/voice

The script prints JSON including:

• ogg_path: send this to Feishu as the voice file (必须在 Workspace 下)
• mp3_path: intermediate file
• voice: selected Edge TTS voice

Sending pattern

After building the file:

1. Send the visible text reply.
2. Send ogg_path with:
   • channel feishu
   • path = generated ogg_path
   • mimeType = audio/ogg
   • asVoice = true

Resource

• scripts/build_feishu_voice.py: deterministic synthesis + ffmpeg conversion for Feishu voice playback.