Audio Model Output

Get audio output from model responses in Agno agents.

Similar to providing audio inputs, you can also get audio outputs from an agent. Take a look at the compatibility matrix to see which models support audio as output.

Audio response modality

The following example demonstrates how some models can directly generate audio as part of their response.

audio_agent.py

from agno.agent import Agent, RunOutput
from agno.models.openai import OpenAIResponses
from agno.utils.audio import write_audio_to_file

agent = Agent(
    model=OpenAIResponses(
        id="gpt-5.2-audio-preview",
        modalities=["text", "audio"],
        audio={"voice": "alloy", "format": "wav"},
    ),
    markdown=True,
)
response: RunOutput = agent.run("Tell me a 5 second scary story")

# Save the response audio to a file
if response.response_audio is not None:
    write_audio_to_file(
        audio=agent.run_response.response_audio.content, filename="tmp/scary_story.wav"
    )

You can find the audio response in the RunOutput.response_audio object.

There is a distinction between audio response modality and generated audio artifacts. When the model responds with audio, it is stored in the RunOutput.response_audio object. The generated audio artifacts are stored in the RunOutput.audio list.

Audio input and Audio output

The following example demonstrates how to provide a combination of audio and text inputs to an agent and obtain both text and audio outputs.

audio_agent.py

import requests
from agno.agent import Agent
from agno.media import Audio
from agno.models.openai import OpenAIResponses
from agno.utils.audio import write_audio_to_file
from rich.pretty import pprint

# Fetch the audio file and convert it to a base64 encoded string
url = "https://openaiassets.blob.core.windows.net/$web/API/docs/audio/alloy.wav"
response = requests.get(url)
response.raise_for_status()
wav_data = response.content

agent = Agent(
    model=OpenAIResponses(
        id="gpt-5.2-audio-preview",
        modalities=["text", "audio"],
        audio={"voice": "sage", "format": "wav"},
    ),
    markdown=True,
)

run_response = agent.run(
    "What's in these recording?",
    audio=[Audio(content=wav_data, format="wav")],
)

if run_response.response_audio is not None:
    pprint(run_response.content)
    write_audio_to_file(
        audio=run_response.response_audio.content, filename="tmp/result.wav"
    )