voice_clone tool reports "corrupted" files when using TTS-generated audio

[mbochneak]

Description

The voice_clone tool returns a "File upload is corrupted" error when attempting to use audio files generated by the text_to_speech tool for voice cloning.

Steps to Reproduce

1. Generate an audio file using text_to_speech
2. Use the generated MP3 file with voice_clone
3. Observe the error message

Error Message

File upload is corrupted. Please ensure it is playable audio.

Questions

I'm wondering if this might be related to:

1. Whether the ElevenLabs API expects original voice recordings rather than TTS-generated audio for cloning
2. The file format/quality requirements for the voice cloning API
3. How the handle_input_file function processes files for upload

Context

This affects the workflow where users might want to clone voices from previously generated TTS audio, though I understand if there are technical limitations that require original voice recordings.

Location

elevenlabs_mcp/server.py, lines 378-395

Would appreciate any guidance on the expected file requirements for voice cloning! Thanks for building such a comprehensive MCP server.