Fal AI STT

The FalSTTService provides high-speed speech-to-text transcription using Fal's Wizper API. It is built as a segmented service, handling audio buffering and speech detection natively for optimal performance.

Installation

To use Fal AI, install the required dependencies:

pip install "piopiy-ai[fal]"

Prerequisites

• A Fal AI API key (Get yours here).
• Set your API key in your environment:

  export FAL_KEY="your_api_key_here"

Configuration

FalSTTService Parameters

Parameter	Type	Default	Description
api_key	str	None	Your Fal API key (falls back to FAL_KEY).
sample_rate	int	None	Audio sample rate in Hz.
params	InputParams	None	Configuration parameters (see below).

InputParams Parameters

Option	Type	Default	Description
language	Language	Language.EN	Language of the audio input.
task	str	"transcribe"	Task to perform (transcribe or translate).
chunk_level	str	"segment"	Level of audio chunking.
version	str	"3"	Version of the Wizper model to use.

Usage

Basic Setup

import os
from piopiy.services.fal.stt import FalSTTService

stt = FalSTTService(
    api_key=os.getenv("FAL_KEY")
)

With Translation Task

import os
from piopiy.services.fal.stt import FalSTTService
from piopiy.transcriptions.language import Language

stt = FalSTTService(
    api_key=os.getenv("FAL_KEY"),
    params=FalSTTService.InputParams(
        language=Language.FR, # Input is French
        task="translate"      # Translate to English
    )
)

Notes

• Segmented Processing: Fal STT uses a segmented approach, meaning it processes chunks of audio after a pause in speech is detected, rather than token-by-token streaming.
• Translation Support: The task="translate" parameter allows for real-time translation from any supported language into English.