OpenClaw Bug Report: TTS Voice Note Delivery Regression in v2026.4.9

Summary

TTS (text-to-speech) voice note delivery stopped working after upgrading to OpenClaw v2026.4.9. The tts tool reports successful audio generation, but voice notes are never delivered to Telegram. STT (speech-to-text) for inbound voice notes is also non-functional.

Environment

• OpenClaw version: 2026.4.9 (updated 2026-04-09)
• Previous working version: Pre-2026.4.9 (TTS worked before this update)
• Channel: Telegram
• OS: Ubuntu (VM in VirtualBox on Intel Mac Mini)
• TTS Provider: ElevenLabs (eleven_multilingual_v2)

Configuration

{
  "messages": {
    "tts": {
      "provider": "elevenlabs",
      "providers": {
        "elevenlabs": {
          "apiKey": "sk_...",
          "baseUrl": "https://api.elevenlabs.io",
          "modelId": "eleven_multilingual_v2",
          "voiceId": “REDACTED"
        }
      }
    }
  },
  "channels": {
    "telegram": {
      "enabled": true,
      "defaultTo": "REDACTED",
      "botToken": "REDACTED"
    }
  },
  "plugins": {
    "entries": {
      "elevenlabs": {
        "enabled": true,
        "config": {}
      }
    }
  }
}

Steps to Reproduce

TTS Issue (Outbound Voice Notes)

1. Send a text message to the assistant via Telegram
2. Assistant calls tts tool with channel: "telegram"
3. Tool returns Generated audio reply
4. Expected: Voice note delivered to user
5. Actual: Only text response received; no voice note

Inbound STT

1. Send a voice note to the assistant via Telegram
2. Expected: Audio transcribed and processed
3. Actual: No transcription available; pdf tool fails with "Expected PDF but got audio/ogg", image tool fails with "Unsupported media type: audio"

Evidence

1. TTS Tool Reports Success

Generated audio reply.

2. No Gateway Activity Logged

Gateway logs show no TTS-related events after tool call:

2026-04-10T10:39:45.453Z error [tools] image failed: Unsupported media type: audio
[NO TTS EVENTS LOGGED HERE]
2026-04-10T10:33:58.570Z info gateway loading configuration…

The gap between 10:33 and 10:39:45 should contain TTS delivery attempts, but nothing appears.

3. Historical Evidence of sendVoice Failure

Earlier logs from ~10:11 UTC show the last sendVoice attempts failing:

2026-04-10T10:11:59.515Z error gateway/channels/telegram telegram sendVoice failed: Network request for 'sendVoice' failed!
2026-04-10T10:11:59.521Z error gateway/channels/telegram telegram final reply failed: HttpError: Network request for 'sendVoice' failed!

After this, no further sendVoice attempts appear in logs.

4. Text Messages Work Fine

2026-04-10T10:21:17.592Z info gateway/channels/telegram telegram sendMessage ok chat=8775399894 message=404

Diagnosis

The TTS audio generation via ElevenLabs API is working (credentials valid, API responsive). However, the tool→channel delivery handoff is broken:

1. Tool generates audio file successfully
2. Tool reports Generated audio reply
3. Audio file is never passed to the gateway's Telegram channel
4. No sendVoice API call is made to Telegram
5. User receives no voice note

This suggests a regression in how the tts tool result is handled by the channel delivery layer in v2026.4.9.

Pre-Update Behavior (Prior to v2026.4.9)

TTS was generating audio but it was corrupted. In the webchat interface, corrupted voice notes appeared as long strings of exclamation marks ("!!!!!!!!!!!!!"). This suggests the audio file was being created with malformed content, and the webchat was displaying the raw binary data.

Key observation: The "!!!!!" corruption appeared when using both Telegram AND webchat simultaneously, suggesting session handling issues contributed to the audio corruption.

Post-Update Behavior (v2026.4.9)

After the v2026.4.9 update, TTS still generates audio and reports success, but no audio file is delivered at all. The behavior shifted from "corrupted delivery" to "no delivery."

Impact

• Critical: Users cannot receive voice note responses (TTS broken)
• Major: Users cannot send voice notes for transcription (STT broken)
• Text messaging remains functional

Additional Notes

• Gateway Telegram channel reconnects successfully after config changes
• Network connectivity to Telegram API is stable (as evidenced by working text messages)
• ElevenLabs API credentials are valid and functional (manual API calls succeed)
• Multi-session factor: Session handling issues suspected when multiple channels (Telegram + webchat) are active simultaneously

Workaround

None available. Manual file attachment via message tool with media path untested but may work as alternative.

Related

• Pre-v2026.4.9: Corrupted audio delivery ("!!!!!!" in webchat)
• v2026.4.9+: No audio delivery at all
• Session handling suspected when multiple channels active