Step-Audio-2

Step-Audio-2#

Step-Audio-2-Mini is a speech-to-speech (STS) model developed by StepFun, featuring an 8B parameter LLM backbone with advanced audio understanding and generation capabilities.

Quickstart#

Start the server:

python -m vox_serve.launch --model stepfun-ai/Step-Audio-2-mini --port 8000

Text-to-Speech#

import requests

response = requests.post("http://localhost:8000/generate", json={
    "text": "Hello, this is Step-Audio speaking!"
})

with open("output.wav", "wb") as f:
    f.write(response.content)

Speech-to-Speech#

import requests

with open("input.wav", "rb") as audio_file:
    response = requests.post(
        "http://localhost:8000/generate",
        files={"audio": audio_file},
        data={"text": "Please respond to my question."}
    )

with open("output.wav", "wb") as f:
    f.write(response.content)

API Parameters#

Parameter

Type

Default

Description

text

string

required

Text prompt or instruction

audio

file

optional

Input audio file for STS mode

audio_path

string

optional

Path to input audio file

streaming

boolean

false

Enable streaming response

Examples#

Text-to-Speech#

import requests

response = requests.post("http://localhost:8000/generate", json={
    "text": "The quick brown fox jumps over the lazy dog."
})

Speech-to-Speech with Audio File#

import requests

# Using file upload
with open("question.wav", "rb") as f:
    response = requests.post(
        "http://localhost:8000/generate",
        files={"audio": f},
        data={"text": "Answer the question in the audio."}
    )

Speech-to-Speech with Audio Path#

import requests

response = requests.post("http://localhost:8000/generate", json={
    "text": "Respond to the audio message.",
    "audio_path": "/path/to/input.wav"
})

Streaming Audio#

import requests

with requests.post(
    "http://localhost:8000/generate",
    json={"text": "Hello world!", "streaming": True},
    stream=True
) as response:
    with open("output.wav", "wb") as f:
        for chunk in response.iter_content(chunk_size=8192):
            f.write(chunk)

Using curl (TTS)#

curl -X POST "http://localhost:8000/generate" \
  -H "Content-Type: application/json" \
  -d '{"text": "Hello from Step-Audio!"}' \
  -o output.wav

Using curl (STS)#

curl -X POST "http://localhost:8000/generate" \
  -F "text=Respond to this audio" \
  -F "audio=@input.wav" \
  -o output.wav

CLI Options#

python -m vox_serve.launch \
  --model stepfun-ai/Step-Audio-2-mini \
  --port 8000 \
  --temperature 0.7 \
  --top_p 0.9

Architecture Notes#

Step-Audio-2 features:

  • 8B parameter Qwen-based backbone

  • Whisper-style audio encoder for input processing

  • Flow matching decoder with HiFT vocoder

  • Support for both text-to-speech and speech-to-speech modes

Contents