openclaw - 💡(How to fix) Fix Telegram forum-topic voice notes reach agent as raw audio while DM transcription/echo works

Summary

On OpenClaw 2026.5.12, Telegram voice notes behave differently between direct chats and forum-topic group chats.

In a Telegram DM, the same gateway transcribes the voice note and sends the configured echoTranscript message.

In a Telegram forum-topic group chat, the voice note reaches the agent as a raw .ogg attachment, but the media-understanding transcript/echo path does not visibly fire. Replying to the voice note with an @mention in the topic still leaves the agent seeing raw audio rather than the transcript.

Environment

• OpenClaw: 2026.5.12 (f066dd2)
• Channel: Telegram
• Group type: Telegram supergroup with forum topics enabled
• Gateway mode: local / loopback
• Audio config:

{
  "tools": {
    "media": {
      "audio": {
        "enabled": true,
        "echoTranscript": true,
        "echoFormat": "📝 \"{transcript}\""
      }
    }
  }
}

Observed behavior

Direct chat

1. Send a Telegram voice note directly to the bot.
2. OpenClaw posts an echo transcript, for example: 📝 "This is a task for all.".
3. Agent receives/uses the machine transcript and responds normally.

Result: transcription and echoTranscript both work.

Forum-topic group chat

1. Send a Telegram voice note inside a forum topic.
2. Mention the bot in a reply to that audio message, because Telegram voice notes do not support normal captions/mentions in the same way.
3. The agent receives a raw audio attachment, for example <media:audio> / .ogg.
4. No visible echoTranscript message appears in the topic.
5. The transcript is not available to the agent in the same way it is in DM.

Result: raw audio reaches the agent, but the media-understanding transcript/echo path does not fire visibly.

Expected behavior

Telegram forum-topic voice notes should go through the same media-understanding transcription path as Telegram DM voice notes, or the docs should clearly state that forum-topic voice notes require a different workflow.

If echoTranscript is enabled, the transcript echo should route to the same forum topic/thread that received the voice note.

Why this matters

For Telegram forum-topic setups, voice notes are a natural input method, but Telegram does not allow clean "voice note + caption mention" usage. If OpenClaw only transcribes DMs reliably, users need to route voice capture through DMs instead of topic-based workflows.

Related issues

This looks related but not identical to:

• #63349
• #53364
• #75084
• #62205
• #54013

Those appear closed or report a different DM/forum direction. This report is specifically: DM works, forum-topic group path does not.