Minute
AI-powered meeting transcription and minute generation, built by i.AI for UK government
Overview
Minute is an open-source tool built by i.AI (the UK Government's AI team in DSIT) that transforms meeting recordings into structured, professional meeting minutes using AI.
Learning Artifact: This is a pre-deployed demonstration environment for learning and exploration, not a production-ready product.
Minute brings together automatic speech transcription and AI-powered minute generation in a single web application. Upload a meeting recording, get an automatic transcript with speaker identification, then generate structured meeting minutes — in minutes, not hours.
The NDX:Try version replaces the upstream Gemini/Azure dependencies with AWS-native services: Amazon Bedrock (Nova Pro) for text generation and AWS Transcribe for speech-to-text.
What you’ll explore
Core Features
| Feature | AWS Service |
|---|---|
| Meeting transcription | AWS Transcribe |
| AI minute generation | Amazon Bedrock (Nova Pro) |
| Speaker identification | AWS Transcribe |
| Multiple meeting templates | Cabinet, Planning Committee, General, and more |
| AI-assisted editing | Amazon Bedrock |
| Word document export | Built-in |
Infrastructure
| Component | AWS Service |
|---|---|
| Compute | AWS Fargate (3 services) |
| Database | Amazon Aurora PostgreSQL |
| Message queues | Amazon SQS (4 queues) |
| File storage | Amazon S3 |
| CDN & HTTPS | Amazon CloudFront |
| Service discovery | AWS Cloud Map |
Getting started
- Request your session — Click “Try this now” above. Your environment will deploy automatically.
- Find your credentials — In the AWS Console, go to CloudFormation → MinuteStack → Outputs tab. Copy the MinuteUrl, BasicAuthUsername, and BasicAuthPassword.
- Log in — Open the MinuteUrl in your browser. Enter the username and password when prompted.
- Upload a recording — Click “New meeting”, then “Upload a file”. Choose an audio file (MP3, WAV, or WebM up to 5GB) and select a template.
- Wait for transcription — AWS Transcribe processes the audio (5-15 minutes depending on length).
- Review your minutes — AI-generated structured minutes appear with meeting overview, attendees, discussion points, key decisions, action items, and next steps.
Why this matters for local government
Reduce administrative burden
Meeting minutes typically take 2-4 hours to produce manually. Minute generates structured minutes in under 15 minutes — freeing officers for higher-value work.
Improve record keeping
AI-generated minutes include citations back to the transcript, structured action items with owners, and consistent formatting across all meeting types.
Multiple meeting formats
Built-in templates for Cabinet meetings, Planning Committees, Care Assessments, and General meetings — plus the ability to create custom templates for your specific needs.
Open source and reusable
Built by i.AI and open-sourced at github.com/i-dot-ai/minute. Already in use across central government — this NDX:Try deployment lets you evaluate it on AWS infrastructure.
Constraints
This is a time-limited evaluation environment provided through the NDX Innovation Sandbox:
- Budget: Fixed allocation per session (sufficient for the walkthrough and extended exploration)
- Duration: Sessions are time-limited — complete your evaluation within the allocated period
- Purpose: Learning and evaluation only — do not upload recordings containing sensitive or classified information
- Data: All data is deleted when the session ends
Important Notes
- Audio formats: MP3, WAV, WebM, and other common formats supported (up to 5GB)
- Transcription time: Varies with recording length — a 30-minute meeting takes approximately 5-10 minutes to transcribe
- LLM model: Uses Amazon Nova Pro via Bedrock — no Anthropic marketplace agreement required
- Authentication: Basic HTTP auth with a randomly generated password (shown in CloudFormation Outputs)
- Speaker identification: AWS Transcribe automatically identifies different speakers, though names must be assigned manually
How this was built
Minute was originally built by i.AI using Terraform, Keycloak, and Google Gemini. The NDX:Try version converts this to AWS CDK with:
- Bedrock adapter replacing Gemini for AI text generation
- AWS Transcribe for speech-to-text (already supported upstream)
- CloudFront basic auth replacing Keycloak for authentication
- Aurora PostgreSQL for the database
- ECS Fargate for all three application services
Source code: github.com/i-dot-ai/minute (upstream) | NDX:Try CDK stack (deployment)
Explore more scenarios
- Council Chatbot — AI-powered chatbot for citizen enquiries using Amazon Bedrock
- Simply Readable — Document translation and Easy Read conversion
- LocalGov Drupal — AI-enhanced content management for councils
- Text to Speech — Convert written content to natural speech with Amazon Polly
Troubleshooting
“Transcription being processed” takes a long time
Transcription time depends on audio length. A 45-minute meeting typically takes 5-10 minutes. The page auto-refreshes when complete.
“There was a problem processing your request” on minute generation
Click “Generate New” and try again. The AI model may occasionally time out on very long transcripts.
“Short meeting detected” message
The transcript must contain at least 200 words for full minute generation. Try with a longer recording.
Cannot access the URL
Ensure you’re using the credentials from the CloudFormation Outputs tab. The password is randomly generated for each deployment.
Support
For help with this scenario or NDX:Try:
- Walkthrough: Step-by-step guide
- Source code: i-dot-ai/minute on GitHub
- NDX Team: Contact via the NDX support channel