Recognition

The Recognition controller converts audio into text (speech recognition). It provides one unified interface across multiple providers.

Supported Providers

Common options:

• openai (cloud transcription)
• keras (offline Whisper models)
• elevenlabs
• speechmatics (batch / real-time via wrapper capabilities)

Parameters

• key_value: API key for providers that require it (e.g., openai, elevenlabs, speechmatics). Not required for keras.
• provider (optional): Provider name (defaults to openai).
• model_name (optional): Used by keras provider (default is whisper_tiny_en).
• model_params (optional): Used by keras to configure the local model.
• input_params: An instance of SpeechRecognitionInput describing the audio source and options.

Example (OpenAI)

from intelli.controller.remote_recognition_model import RemoteRecognitionModel
from intelli.model.input.text_recognition_input import SpeechRecognitionInput

model = RemoteRecognitionModel(
    key_value="your_openai_api_key",
    provider="openai",
)

input_params = SpeechRecognitionInput(
    audio_file_path="path/to/audio.mp3",
    language="en",
    model="whisper-1",
)

text = model.recognize_speech(input_params)
print(text)

Example (Offline / Keras Whisper)

Install offline extras:

pip install "intelli[offline]"

from intelli.controller.remote_recognition_model import RemoteRecognitionModel
from intelli.model.input.text_recognition_input import SpeechRecognitionInput

model = RemoteRecognitionModel(
    provider="keras",
    model_name="whisper_tiny_en",
)

input_params = SpeechRecognitionInput(
    audio_file_path="path/to/audio.mp3",
    language="<|en|>",
)

text = model.recognize_speech(input_params)
print(text)

---

Speechmatics (extra dependencies)

To use Speechmatics, install the speech extra:

pip install "intelli[speech]"

This extra corresponds to the speech section in setup.py and includes packages like: speechmatics-batch, speechmatics-rt, websockets, librosa, soundfile, numpy, and openai>=2.5.0.

Minimal usage:

from intelli.controller.remote_recognition_model import RemoteRecognitionModel
from intelli.model.input.text_recognition_input import SpeechRecognitionInput

model = RemoteRecognitionModel(
    key_value="your_speechmatics_api_key",
    provider="speechmatics",
)

input_params = SpeechRecognitionInput(
    audio_file_path="path/to/audio.wav",
    language="en",
)

text = model.recognize_speech(input_params)
print(text)