Introducing “Transcribe Call” in AI Studio for HubSpot

Overview

The “Transcribe Call” workflow action in AI Studio for HubSpot automatically transcribes recorded calls and returns a text transcript within your HubSpot workflow. Designed for HubSpot's Call-based Workflows, it enables efficient transcription and further AI processing of the call transcript with leading AI models (with Send to AI workflow action). 

Key Features & Benefits

1. Multilingual Transcription

• Uses OpenAI Whisper to detect and transcribe more than 50 languages.

• Automatically handles language switches during the same call.

2. Pause Removal & Audio Optimization

• Trims silent or pause segments to reduce token usage and improve clarity.

3. Native HubSpot Integration

• No separate platform needed—add “Transcribe Call” as an action in your workflows.

4. Follow-Up Possibilities

• Integrate with “Send to AI” to generate summaries, draft emails, extract keywords, or perform sentiment analysis

How to get started?

Register at ai.resonatehq.com and connect your HubSpot account (portal)

Setting Up a Call-Based Workflow

1. Create or Edit a Workflow

• Go to Automation > Workflows in HubSpot and select Create workflow (or edit an existing one).

• Choose Call-based (or another object type if desired).

2. Define Enrollment Criteria

• Example: “Recording URL is known,” ensuring only calls with valid audio links enter the workflow.

3. Add “Transcribe Call”

• Click + to add a new action.

• Select “AI Studio for Workflows”, then choose “Transcribe Call.”

How It Works

1. Audio Retrieval

• AI Studio accesses the audio file from the provided URL.

2. Pause Removal

• Silent segments are trimmed to reduce token usage.

3. OpenAI Whisper Transcription

• Whisper processes the audio, detecting language automatically.

4. Transcript Output

• The completed text transcript is made available as a workflow output variable (Transcription).

Using the Transcription Output

Scenario 1: Save Transcript to Call Notes

1. Add an “Edit record” action after “Transcribe Call.”

2. Set Call notes (or any other text property) to the Transcription output.

Scenario 2: Send to AI for Analysis

1. Insert a “Send to AI” step (also from AI Studio) after “Transcribe Call.”

2. Choose your AI model (OpenAI GPT-4o mini, Google Gemini Pro, etc.).

3. Reference the output of Transcription action in the prompt to generate summaries, follow-up emails, or sentiment analysis.

Cost & Token Usage

• 500 tokens per minute of audio are used for transcription.

• Pause removal lowers token usage by skipping silent sections.

• Track AI usage credits in AI Studio’s dashboard at ai.resonatehq.com

Language Detection & Multilingual Support

• Automatic Detection: No manual switching required for different languages.

• 50+ Languages: English, Spanish, French, Japanese, Arabic, Chinese, etc.

• Handles language switches within a single call whenever possible.