GPTfy Voice Recorder feature allows users to record a voice note, which is then sent to the AI for transcription. The sales team can extensively utilize this feature to document pre-sales conversations.
To ensure this solution works properly, the following settings must be enabled within the AI Settings under the Voice Configuration tab.
Prerequisites for Voice Recorder Solution:
- Google Chirp AI model/ Kensho must be activated. Refer this article for configurations and settings.
- A custom notification, ‘Notify_Event_created’, must be created and enabled for both desktop and mobile notifications.
- Notification Delivery Settings should be done as shown below:-
- GPTfy_Notification_Flow.zip needed to be imported using the workbench.
Set up the voice recorder
To set up the voice recorder solution, follow these steps:
- Go to any record where you want to run the GPTfy Voice and click on the Edit page. Drag and Drop the Voice Recorder Component and Add the voice recorder component to the record’s Lightning page by dragging and dropping it into the desired location.
- Add the Audio listener component to the page layout.
- In the AI settings, navigate to the Voice Configuration tab and add the connection details.
- Once all three models of Kensho are activated in the AI Model, add them to the Voice Configuration settings under AI settings.
- AI Setting Voice Configuration:
- Duration (Minutes): Add a value in minutes within the Voice Configuration settings. The recorder will function for recordings for the mentioned duration. The maximum limit is 3.
- Audio Processing Model: Choose the available voice model in the Audio File Upload Model.
- Audio polling model: Add the related Audio polling model.
- Polling Timeframe (Minutes): The time interval (in minutes) between each poll attempt. If the AI fails to generate the transcript (due to a timeout, failure, or error), the process retries after the defined polling interval.
- Polling Counter: If the transcript is successfully generated, the process stops. The polling counter is the maximum number of attempts before the process stops.
- Enable the Solution:
- Create a prompt on the Content Version object with the transcript field selected in Field Mapping.
- Add Action Type as Invoke Flow in Prompt Actions with a Flow named GPTfy Notification Flow
Note: GPTfy Notification Flow should be present in the org for notifications and event creation
- Watch this video recording for Flow Creation.
- Click on Save and Activate the Prompt.
- Add this prompt in the Voice Configuration details section and the Related To Object field from a picklist.
- Once the setup is complete, users can record their voice for a duration specified by the configured time limit in minutes. After recording, users can choose to either delete or upload the recording.
Click the upload button. Then, the audio is transcribed into text and returned to Salesforce.
- An MP3 audio file should be created and attached to the record where the voice recorder was used.
- Transcript field: It should be added to the content version page layout.
- When the transcript is created the user gets a notification on desktop/Mobile.
- Click on this notification and the user will redirect to the event page which is created on that particular record where the Voice was recorded. This event will have all the details including the description which consists of the transcript of the recorded voice.
