Voxjar lets you bring your own transcripts.
This is usually not necessary, since Voxjar uses state of the art speech to text machine learning models.
But we understand that you might prefer to use your own transcripts for any number of reasons.
This guide will show you how to format your transcripts to work with Voxjar, including our AI evaluator, and how to get your transcripts into the system via our integrations.
Formatting Your Transcripts to Match
Data formatting can be a pain. So instead of forcing your transcripts into a custom data structure, you just need to send it as a block of text (a string).
However you text is sent to Voxjar, is exactly how we'll display it on the website, and how we'll send it to the AI.
You can experiment with different layouts (whitespace, labels, etc.) but a good place to start is to separate chunks of the transcript with newlines/whitespace and label each chunk with the speaker and timestamps.
Here is an example:
Agent <timestamp hh:mm:ss>
Transcript chunk
Customer <hh:mm:ss>
Transcript chunk
If your transcript is not sent as a string, or is not sent at all, Voxjar will automatically transcribe your audio files.
Integrations that support Bring Your Own Transcripts
Currently, Voxjar supports bring-your-own-transcripts via our cloud storage and FTP integrations.
You'll just need to include the transcripts in your metadata files.
There are four integrations to look into:
Check out those help articles for detailed guides.
Map your Transcripts as Metadata
Make sure your transcript is added to your metadata files.
Designate a column in your csv file or a key in your json file and include your transcript as the value.
Then when you connection your integration, map the transcript column or key to the "transcript" metadata field.
CSV File Mappings
Find your transcript column and select the "transcript" metadata field from the dropdown.
JSON File Mappings
Find your transcript key and select the "transcript" metadata field from the dropdown.
