Eagle Speaker Recognition Demos

Made in Vancouver, Canada by Picovoice

Eagle

Eagle is an on-device speaker recognition engine. Eagle is:

• Private; All voice processing runs locally.
• Cross-Platform:
  • Linux (x86_64), macOS (x86_64, arm64), Windows (x86_64, arm64)
  • Android and iOS
  • Chrome, Safari, Firefox, and Edge
  • Raspberry Pi (3, 4, 5)

Compatibility

• Python 3.9+
• Runs on Linux (x86_64), macOS (x86_64, arm64), Windows (x86_64, arm64), and Raspberry Pi (3, 4, 5).

Installation

pip3 install pveagledemo

AccessKey

Eagle requires a valid Picovoice AccessKey at initialization. AccessKey acts as your credentials when using Eagle SDKs. You can get your AccessKey for free. Make sure to keep your AccessKey secret. Signup or Login to Picovoice Console to get your AccessKey.

Overview

Eagle consists of two distinct steps: Enrollment and Recognition. In the enrollment step, Eagle analyzes a series of utterances from a particular speaker to learn their unique voiceprint. This step results in an EagleProfile object, which can be stored and utilized during inference. During the Recognition step, Eagle compares the incoming frames of audio to the voiceprints of all enrolled speakers in real-time to determine the similarity between them.

Microphone Demo

The microphone demo captures audio input from a microphone that is connected. To run the demo, use the following command in the terminal:

eagle_demo_mic {enroll, test} --access_key ${ACCESS_KEY} ...

Replace ${ACCESS_KEY} with yours obtained from Picovoice Console.

The commands enroll and test are used to create a speaker profile and perform speaker recognition, respectively. Detailed explanations of these commands will be provided in their respective sections.

Furthermore, the demo offers optional arguments, which can be accessed by utilizing the --help argument. By doing so, you will receive a comprehensive listing of the available arguments along with their corresponding descriptions.

eagle_demo_mic --help

Speaker Enrollment

If the demo is executed in the enrollment mode by using the enroll command, it will initiate the enrollment process using the audio captured from the microphone. It will display the progress percentage in the terminal until it reaches 100%. Once completed, it will save the profile of the enrolled speaker to the disk.

eagle_demo_mic enroll --access_key ${ACCESS_KEY} --output_profile_path ${OUTPUT_PROFILE_PATH}

Replace ${OUTPUT_PROFILE_PATH} with the absolute path where the generated profile should be written.

Speaker Recognition

Once the speaker profile for all speakers are created, the demo can be run in the test mode by running the following command:

eagle_demo_mic test --access_key ${ACCESS_KEY} --input_profile_paths ${INPUT_PROFILE_PATH_1 ...}

In this mode, you can include multiple speaker profiles by specifying them with the --input_profile_paths option. Eagle will assess and provide a distinct score for each profile, which will be displayed in the terminal.

Note: the demo uses the default input audio device on your machine. To choose a different device, use the --show_audio_devices option to see a list of available devices and their indexes, then run the above commands with the --audio_device_index option:

eagle_demo_mic enroll --access_key ${ACCESS_KEY} --output_profile_path ${OUTPUT_PROFILE_PATH} --audio_device_index 2

File Demo

Similar to the mic demo, the file demo can be run in two modes: enroll and test

eagle_demo_file {enroll,test} --access_key ${ACCESS_KEY} ...

Replace ${ACCESS_KEY} with yours obtained from Picovoice Console.

The commands enroll and test are used to create a speaker profile and perform speaker recognition, respectively, and will be discussed in detail in their respective sections.

To view the optional arguments for the demo, use the --help argument. This will display a list of available arguments and their descriptions.

eagle_demo_file --help

Speaker Enrollment

To run the demo in enroll mode, you need two additional input arguments along with the AccessKey.

eagle_demo_file enroll --access_key ${ACCESS_KEY} \
  --output_profile_path ${OUTPUT_PROFILE_PATH} --enroll_audio_paths ${ENROLL_AUDIO_PATH_1 ...} 

In this command, {ENROLL_AUDIO_PATH_1 ...} represents the absolute paths to the enroll audio files. If multiple files are provided, Eagle will process all of them. Once the specified files are processed, the demo will generate a profile at ${OUTPUT_PROFILE_PATH}.

Speaker Recognition

The file demo requires a test audio and one or more speaker profiles that were created during the enrollment step.

To run the demo, use the following command in the console:

eagle_demo_file test --access_key ${ACCESS_KEY} \
  --input_profile_paths ${INPUT_PROFILE_PATH_1 ...} --test_audio_path ${TEST_AUDIO_PATH}

The demo will display the result for each enrolled speaker in the terminal.

Optionally, you can also generate a .csv file for further analysis by including the --csv_output_path parameter:

eagle_demo_file test --access_key ${ACCESS_KEY} \
  --input_profile_paths ${INPUT_PROFILE_PATH_1 ...} --test_audio_path ${TEST_AUDIO_PATH} \
  --csv_output_path ${CSV_OUTPUT_PATH}