Three specific playable Telegram voice messages consistently fail with HTTP 400 and link to missing attachments

[Ingressia228]

Description

Three specific Telegram voice messages consistently fail when processed by the plugin.

For each affected message, the plugin displays:

Voice transcription failed (HTTP 400). The attachment was still saved.

An audio embed is then added to the Obsidian note. When opening it, Obsidian reports:

The note "<attachment path>" does not exist. Click to create it.

The three affected voice messages play normally in Telegram.

Other voice messages sent immediately before and after them are saved and transcribed normally.

Environment

• OS: macOS
• Plugin: Vault Telegram Bridge 0.3.7
• Previous transcription model: whisper-1
• Current transcription model: gpt-transcribe

Only the transcription model was changed. The transcription provider and API endpoint were not changed.

The affected messages and the successfully processed messages immediately before and after them were all tested using gpt-transcribe.

The plugin displayed only the HTTP 400 status. No transcription API response body was shown.

Retesting

I forwarded the same three affected voice messages multiple times.

Each message was forwarded in both available ways:

• with the original sender attribution visible;
• with the original sender attribution hidden.

The same behavior occurred in every case:

1. Transcription failed with HTTP 400.
2. The plugin reported that the attachment was saved.
3. An audio embed was inserted into the Obsidian note.
4. Obsidian could not find the attachment at the linked path.

The same three voice messages failed on every forwarding attempt. Other voice messages continued to work.

Privacy

The affected voice messages are not attached to this report.

Sender information, filenames, local paths, dates, timestamps, chat identifiers, credentials, and message contents are not included.