Full details on github: https://github.com/dan-mirescu/SpeechRecognizer
SpeechRecognizer is a functional demo application for Windows designed for usage inside VR virtual desktop.
Note: the application can also be used outside VR virtual desktop, however it is built with VR in mind.
While speech recognition already exists in Windows, it did not exist for my native language. Additionally, I liked the quality of speech recognition that the Speech-to-Text service from Google offers. Therefore, I decided to create a desktop application that will help me to type text more easily while using VR virtual desktop.
- Only two languages are available: Romanian and English (US). New languages can be easily added on request or by pulling the code yourself and adding them.
- Visual design is not the most appealing but I aimed for functionality / for an MVP
The speech recognition service is offered by Google and it is not 100% free. You need to configure the Speech-to-Text service in Google Cloud with a Google account and use the credentials generated from there with this application. Generally, Google offers a certain amount of free minutes of service per month and low prices per minute after the free minutes elapse. See the pricing details here: https://cloud.google.com/speech-to-text/pricing
Requirements: Google account and Google cloud project with billing enabled
Download the application from the Releases section on Github.
Before opening the application, you need to configure the Google Speech-To-Text service. This will be a one-time operation as your service credentials will be saved in your user profile folder.
- Go to https://cloud.google.com/speech-to-text/docs/before-you-begin and follow the steps until “Set your authentication environment variable”. At this point you should have a JSON file containing the necessary service credentials.
- Open the application and browse to the JSON file or drag and drop it on top of the window. Then, click on “Use provided file”.
After you have set up your service credentials you can open the application. When the application opens, a red “microphone” button is displayed on top of all the windows.
Right click the button for settings:
- ro-RO, en-US: the language to be used for the speech recognition. Note: the application is in a functional demo stage and these are the available languages. New languages can be easily added on request or by pulling the code yourself and adding them.
- Auto send keys: whether the recognized text to be auto-typed from the keyboard when the speech recognition ends
- Auto copy to clipb: whether the recognized text should be auto-placed in the clipboard when the speech recognition ends
The settings are applied instantly when changing them.
The usual flow:
- Select any kind of text input area in Windows
- Click the application “microphone” button
- Wait until the message “Speak now” is displayed
- When the application detects that you stopped, it will show you the recognized text and auto-fill it in the select text input area. Auto-fill happens if “Auto send keys” is checked in the settings.
More info on the application’s github page: https://github.com/dan-mirescu/SpeechRecognizer