Skip to content
llms.py logo llms.py

Audio Support

llms.py supports audio file inputs through audio-capable models, enabling transcription, summarization, and analysis of audio content.

CLI Usage

Section titled “CLI Usage”
Terminal window
# Use default audio template (transcribe)
llms --audio ./recording.mp3


# With custom prompt
llms --audio ./meeting.wav "Summarize this meeting"


# Remote audio URL
llms --audio https://example.org/podcast.mp3 "Key points?"


# With specific audio model
llms -m gpt-4o-audio-preview --audio interview.mp3 "Extract topics"
llms -m gemini-2.5-flash --audio talk.mp3 "Transcribe"


# Combined with system prompt
llms -s "You're a transcription specialist" --audio talk.mp3 "Detailed transcript"

Web UI

Section titled “Web UI”
  1. Click the audio attachment icon
  2. Select an audio file
  3. Add your prompt
  4. Send

Supported Formats

Section titled “Supported Formats”
  • MP3
  • WAV
  • M4A (depending on provider)

Audio Sources

Section titled “Audio Sources”
  • Local files: Absolute or relative paths
  • Remote URLs: HTTP/HTTPS URLs (automatically downloaded)
  • Base64 Data: Base64-encoded audio

Audio-Capable Models

Section titled “Audio-Capable Models”

Popular models that support audio processing:

  • OpenAI: gpt-4o-audio-preview
  • Google: gemini-2.5-pro, gemini-2.5-flash, gemini-2.5-flash-lite

Custom Templates

Section titled “Custom Templates”

You can create custom request templates for audio processing:

Audio Template Example

Section titled “Audio Template Example”
{
  "model": "gpt-4o-audio-preview",
  "messages": [
    {
      "role": "user",
      "content": [
        {
          "type": "input_audio",
          "input_audio": {"data": "", "format": "mp3"}
        },
        {
          "type": "text",
          "text": "Please transcribe this audio"
        }
      ]
    }
  ]
}

Usage:

Terminal window
llms --chat audio-request.json --audio recording.mp3

Next Steps

Section titled “Next Steps”