For better speech recognition quality, you can specify the language of the audio input. If not specified, the system will attempt to automatically detect the language.

​
Using the Fish Audio SDK

First, make sure you have the Fish Audio SDK installed. You can install it from GitHub or PyPI.

​
Example Usage

from fish_audio_sdk import Session, ASRRequest

session = Session("your_api_key")

# Read the audio file
with open("input_audio.mp3", "rb") as audio_file:
    audio_data = audio_file.read()

# Option 1: Without specifying language (auto-detect)
response = session.asr(ASRRequest(audio=audio_data))

# Option 2: Specifying the language
response = session.asr(ASRRequest(audio=audio_data, language="en"))

# Option 3: With precise timestamps (may increase latency for short audio)
response = session.asr(ASRRequest(audio=audio_data, ignore_timestamps=False))

print(f"Transcribed text: {response.text}")
print(f"Audio duration: {response.duration} seconds")

for segment in response.segments:
    print(f"Segment: {segment.text}")
    print(f"Start time: {segment.start}, End time: {segment.end}")
This example demonstrates three ways to use the Speech-to-Text API:
  1. Without specifying a language: The system will attempt to auto-detect the language.
  2. Specifying the language: You can provide the language code (e.g., “en” for English) for potentially better recognition.
  3. With precise timestamps: By setting ignore_timestamps=False, you can get more accurate timing information for each segment. Note that this may increase latency for short audio files.
The ignore_timestamps parameter is set to True by default. This reduces latency for short audio

​
Raw API Usage

If you prefer to use the raw API instead of the SDK, you can use the MessagePack API as described below.

​
Endpoint Details

​
Example Usage

import httpx
import ormsgpack

# Read the audio file
with open("input_audio.mp3", "rb") as audio_file:
    audio_data = audio_file.read()

# Prepare the request data
request_data = {
    "audio": audio_data,
    "language": "en",  # Optional: specify the language
    "ignore_timestamps": False  # Optional: set to True to ignore precise timestamps
}

# Send the request
with httpx.Client() as client:
    response = client.post(
        "https://api.fish.audio/v1/asr",
        headers={
            "Authorization": "Bearer YOUR_API_KEY",
            "Content-Type": "application/msgpack",
        },
        content=ormsgpack.packb(request_data),
    )

# Parse the response
result = response.json()

print(f"Transcribed text: {result['text']}")
print(f"Audio duration: {result['duration']} seconds")

for segment in result['segments']:
    print(f"Segment: {segment['text']}")
    print(f"Start time: {segment['start']}, End time: {segment['end']}")
This example shows how to use the raw API with MessagePack serialization. You can also use multipart/form-data by changing the Content-Type header and adjusting the request data format accordingly. Make sure to replace "YOUR_API_KEY" with your actual API key, and adjust the file paths as needed.