StepAudio 2.5 Chat

​

Specifications

​

Capabilities

• Paralinguistic understanding: For audio input, the model interprets vocal cues such as intonation, pacing, hesitation, and laughter in addition to the spoken words, capturing a speaker’s emotional state and intent that a text transcript does not represent.
• Contextual, emotionally aware dialogue: The model resolves complex or implicit meaning across turns and factors the inferred tone and context into its replies, rather than responding only to the literal words.
• Persona and style control: System instructions define the assistant’s personality, speaking style, and behavioral boundaries at a fine-grained level, maintaining a consistent voice throughout a conversation without preset templates.
• Unified text and audio input: Text and audio are handled through the same Chat Completion endpoint, so an application can combine typed and spoken turns within a single conversation.

​

Common use cases

• Voice assistants and conversational agents: Accept spoken input and return text that the application displays or passes to a text-to-speech model for a spoken reply.
• Call and meeting analysis: Summarize recordings, extract action items, and answer questions about what was discussed.
• Voice message triage: Classify incoming audio by intent and urgency and route it to the appropriate team or workflow without manual review.

​

Quick Start

​

Set up your environment

export STEPFUN_API_KEY="your-api-key"

​

Text in, text out

curl https://api.stepfun.ai/v1/chat/completions \
  -H "Authorization: Bearer $STEPFUN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "stepaudio-2.5-chat",
    "modalities": ["text"],
    "messages": [
      {"role": "system", "content": "You are a friendly, empathetic customer-support assistant for an online store. Acknowledge how the customer feels, then help. Keep replies under three sentences."},
      {"role": "user", "content": "I ordered a jacket two weeks ago and it still has not arrived. This is really frustrating."}
    ]
  }'

​

Audio in, text out

import os
import base64
from openai import OpenAI

client = OpenAI(api_key=os.environ["STEPFUN_API_KEY"], base_url="https://api.stepfun.ai/v1")

with open("customer-message.wav", "rb") as f:
    audio_b64 = base64.b64encode(f.read()).decode()

response = client.chat.completions.create(
    model="stepaudio-2.5-chat",
    modalities=["text"],
    messages=[
        {"role": "system", "content": "You are a friendly, empathetic customer-support assistant for an online store. Acknowledge how the customer feels, then help. Keep replies under three sentences."},
        {"role": "user", "content": [
            {"type": "text", "text": "Here is a voice message from a customer. Please reply."},
            {"type": "input_audio", "input_audio": {"data": f"data:audio/wav;base64,{audio_b64}"}},
        ]},
    ],
)

print(response.choices[0].message.content)

​

Pricing

​

Related Resources