Dispatch AI
Speak to our AI dispatcher over the radio to manage CAD status, dispatch calls, run record lookups, and more.
This feature is available to select communities in our beta testing group and is not yet available to the public.
This feature requires Sonoran One, which includes the pro version of Sonoran Radio, CAD, and more.
Setup
Copy Sonoran CAD API Information
Dispatch AI requires a Sonoran CAD community ID, API key, and server ID.
The community ID and API key are located in the Admin panel > Advanced > In-Game Integration > Web API.
The server ID can be found in the Admin panel > Customization > Servers. The default server ID is 1.
Link CAD In Radio
In the radio Customization panel > Info > Select a server (or create one) and click on the Sonoran CAD icon. This will open a popup modal to enter in your CAD's community ID, API key, and server ID.
Usage
Selecting Your Wake Word
When pressing your push-to-talk key, a local AI listens to your microphone feed for the selected wake word. Once the wake word is heard, it sends all microphone traffic to the external AI for processing.
Customizing your wake word is available in the radio settings. Communities can even create their own, custom wake words.
Prompting the AI
All push-to-talk transmissions must start with the wake word in the first three seconds, or it will be considered a transmission to another unit.
Example
If your wake word is Dispatch, then your transmissions to the AI would be as follows:
<wake word>, <optional: my unit number>, <action to do> Dispatch, A-10, mark me as available. Dispatch, A-10, run a lookup on the license plate ABC123
AI Commands & Features
CAD Functionality
Radio Functionality
Coming soon!
Creating Custom Wake Words
Your wake word is the phrase that tells the AI you’re addressing it, not another radio user, similar to “Hey Siri,” “Okay Google,” or “Alexa.”
Sonoran Radio includes several built-in wake word options, and you can also train a custom wake word tailored to your community. Training is straightforward, though processing can take up to an hour.
1. Build the Model
1A. Wake Word Pronunciation
Open the Google Colab training environment. In the first section, enter your custom target_word, then click the Play button. The first run may take 30–60 seconds (or longer) to prepare the audio.
The webpage will play your custom wake word using its current pronunciation. Make sure the audio matches exactly how you intend to say the wake word. If it sounds incorrect, adjust the spelling to improve pronunciation.
Example: Sonoran → suh_nar_rohn
1B. Running the Model Training
Once you’re satisfied with the wake word’s pronunciation, select Run All at the top of the page or go to Runtime → Run All.
Training may take an hour or longer to complete. When finished, two files will automatically download to your computer.
2. Upload the Model
Now that your model files are ready, it's time to upload them to Sonoran Radio.
This feature is coming soon!
3. Select the Model
Users in the community can now select the custom wake word in their settings. Additionally, you can configure the default wake word for new users in your community.
Last updated
Was this helpful?