Transcript & AI Tools
Recording Transcript
You can get a transcipt of a video or an audio call. The exported file we offer is a .json file that contains an array of sentences along with the time they were spoken. To enable transcript support you will need to setup a transcript provider.
Currently we support the transcript providers below.
- OpenAI
Google
In order to setup Google as a transcript provider, two configuration files are required. One storage configuration file and one speech-to-text configuration file.
Once you have those two files and a storage bucket, start the configuration flow below.
OpenAI
OpenAI as a transcript provider requires an OpenAI API Key and optionally an Organization and Project configured in the OpenAI dashboard.
We assume you have already configured a storage provider to store the recordings.
Once you have the required keys, start the configuration flow below.
Configuration flow
The configuration of a provider consists of the steps below.
- The fist step of the flow is to upload those files.
- The second step of the flow is to select one or more languages that should be available for transcript. You can only pick one language to be transcribed per transcription request.
- The last step of the process is a test page. Click the button and start speaking in the language you selected. Click the button again to stop and start a transcription request. It will take some time for the transcription to finish. Once ready a link with a file will appear. Download the file to see the trascript results.
- Click "Save" to complete the flow.
Tools
Once you have succesfully configured a provider, you have access to the AI tools that provider supports. Currently Only OpenAI supports the advanced tools.
Models
You have the flexibility to choose which OpenAI model will be used for generating transcripts, executing LLM commands such as summarization and sentiment analysis, and performing vision-based analysis. This allows you to tailor the AI’s performance and accuracy to best fit your specific needs and use cases.
Custom prompts
Admins have the option to enable an additional feature for supervisors that allows them to analyze transcripts using a custom prompt. This means you can define a specific prompt tailored to your business needs, ensuring that the AI focuses on the most relevant aspects of the conversation. Furthermore, you can fine-tune and adjust the prompt over time to refine the analysis, making it more precise and aligned with evolving requirements.
Vision (Image analysis)
You have the option to enable the ‘Vision’ tool, which allows agents to capture a snapshot of the customer’s video feed and analyze it in real time using a custom prompt. This tool leverages AI-powered vision analysis to extract insights from the image, making it particularly useful in various customer service and support scenarios.
For example, in a tech support scenario, an agent can take a snapshot of a malfunctioning device shown by the customer and use AI to identify error messages, diagnose hardware issues, or provide troubleshooting steps. In insurance claims processing, an agent can capture an image of a damaged vehicle or property and receive an instant assessment based on predefined criteria. Similarly, in retail customer support, an agent assisting with product-related queries can analyze an image of a received item to verify defects, confirm product authenticity, or provide step-by-step setup guidance.
By allowing agents to use a custom prompt, businesses can tailor the analysis to their specific industry needs, ensuring accurate and context-aware AI-driven insights that enhance customer interactions.
Request a transcript
Once the provider is setup you can go to any interaction that has a recording and request for a transcript. Once the transcript is ready you will be able to chose from the AI tools available for that provider.
Transcript is available once a provider is configured.
