TaskScribe – Local AI Voice Note Assistant

This is a submission for the Hacktoberfest Weekend Challenge: Build for a Friend.

TaskScribe: Turning Voice Notes into Actionable Tasks

My friend often gets useful ideas, reminders, and small tasks while studying, travelling, or doing daily work. Typing every thought immediately is not always convenient, so voice notes are the fastest way to save them.

But there was one problem: voice notes can easily become forgotten audio files. Finding an important reminder later means listening to the recording again, and turning it into an actual task takes extra time.

I built TaskScribe to solve this problem.

The friend’s problem

My friend needed a quick and simple way to capture thoughts without opening multiple apps or typing long notes.

The main challenges were:

• Important ideas were saved as unorganized voice notes.
• It was difficult to search or revisit a specific point from an audio recording.
• Tasks mentioned in voice notes could be forgotten.
• Typing notes immediately was inconvenient.
• Privacy matters when a person records personal reminders, study plans, or work ideas.

The goal was to make voice notes more useful: record a thought once, then turn it into understandable and actionable information.

What I built

TaskScribe is a local AI-powered voice note assistant.

It helps users record or provide a voice note, convert that audio into text, and organize the important information into useful notes and tasks.

Instead of replaying an entire voice recording, a user can quickly review the transcription and see what action is needed.

How TaskScribe works

1. The user records or uploads a voice note.
2. TaskScribe processes the audio.
3. The application creates a text transcription.
4. AI helps organize the raw spoken content into readable notes.
5. Important action items can be identified from the note.
6. The user can review the result in one simple interface.

Project demo

🎥 Demo video: https://youtu.be/LZ00nihvOxM

💻 GitHub repository: https://github.com/Gunjan-2007/TaskScribe

Screenshot

Below is a screenshot of TaskScribe in action:

Technology used

TaskScribe was built using:

• Python for the core application logic.
• Streamlit for the interactive web interface.
• Google Gemma as the open-weight AI model for analyzing transcribed voice notes, generating concise summaries, and extracting actionable tasks.
• Audio input and voice-note handling for capturing spoken thoughts.
• Git and GitHub for version control and project hosting.

Best Use of Gemma

TaskScribe uses Google Gemma as the open-weight AI model at the core of the application.

After a voice note is converted into text, Gemma analyzes the transcription and helps transform unstructured spoken content into a clear summary and actionable tasks.

For example, a user might say:

“I need to finish the TaskScribe README tonight, email the project link to my mentor tomorrow, and remember to prepare the demo video.”

TaskScribe can use Gemma to understand this voice note and organize it into useful output:

• Finish the TaskScribe README tonight.
• Email the project link to the mentor tomorrow.
• Prepare the demo video.

Using Gemma makes TaskScribe more useful than a basic voice recorder because it does not only store audio or display a raw transcription. It helps the user understand what matters and identify what needs to be done next.

Gemma is important to this project because it enables the AI-powered note understanding, summarization, and task extraction workflow. This makes it possible to turn a quick spoken thought into organized and actionable information.

How open-source AI helps

Open-source AI is an important part of TaskScribe because it makes it possible to build useful AI-powered tools with more flexibility and control.

For a voice-note assistant, privacy is especially important. Voice notes may include personal plans, reminders, ideas, or work-related information. A local or open approach can help users have more control over their data instead of unnecessarily relying on a third-party service.

TaskScribe uses Google Gemma, an open-weight AI model, to understand transcribed voice notes and convert them into useful summaries and actionable tasks.

Open-source AI also makes TaskScribe easier to improve in the future. The project can be adapted for different languages, note formats, accessibility needs, and user workflows.

What I learned

While building TaskScribe, I learned how to:

• Build an AI-powered application around a practical everyday problem.
• Work with audio input and voice-note workflows.
• Convert unstructured voice input into more organized information.
• Build a simple interface for an AI feature.
• Document and publish a project using GitHub.
• Think about privacy when creating voice-based AI tools.

Future improvements

I would like to improve TaskScribe further by adding:

• Better support for different accents and languages.
• Task priorities and due-date detection.
• Export options for notes and tasks.
• Searchable note history.
• Better mobile usability.
• More customization for how notes and tasks are organized.

Built for a friend

TaskScribe started with a simple idea: useful thoughts should not be lost just because typing is inconvenient.

By turning voice notes into organized text and actionable tasks, TaskScribe helps my friend capture ideas quickly and return to them later with less effort.

Thank you for reading!