Hey HN, Tony here, I built an on-device AI transcription assistant for professionals who require full data privacy, such as legal and medical users. Everything is offline and transcripts are encrypted in SQLLite database. The audio files and transcripts never leave your device so it fully cloud-free service. My goal is to provide an intuitive, easy-to-use app with unlimited transcriptions and no caps.
I used the following tech stacks to build it: Python (Whisper models), Avalonia UI (frontend), C# (backend), and SQLite (database).
Here's some of the key features:
Speaker Identification: Automatically labels speakers (SPEAKER 01, SPEAKER 02) and allows you to update them with real names.
Whisper Model Transcription: Supports 98 languages and multiple audio/video formats.
Real-Time Transcription: Transcribe audio with your device microphone or system speakers which is ideal for meetings, legal depositions, or court proceedings.
Bulk Transcription: Queue multiple files for batch processing, useful for podcasts or large audio collections.
Proofreading & Editing Tools: Integrated media player and editor for quick corrections or scoping.
AI Models: Choose Intermediate, Advanced, or Expert to balance speed and accuracy.
GPU Acceleration: Boost transcription performance with NVIDIA GPUs.
Export Options: Save transcripts as Word, PDF, or plain text.
I used the following tech stacks to build it: Python (Whisper models), Avalonia UI (frontend), C# (backend), and SQLite (database).
Here's some of the key features:
Speaker Identification: Automatically labels speakers (SPEAKER 01, SPEAKER 02) and allows you to update them with real names. Whisper Model Transcription: Supports 98 languages and multiple audio/video formats. Real-Time Transcription: Transcribe audio with your device microphone or system speakers which is ideal for meetings, legal depositions, or court proceedings. Bulk Transcription: Queue multiple files for batch processing, useful for podcasts or large audio collections. Proofreading & Editing Tools: Integrated media player and editor for quick corrections or scoping. AI Models: Choose Intermediate, Advanced, or Expert to balance speed and accuracy. GPU Acceleration: Boost transcription performance with NVIDIA GPUs. Export Options: Save transcripts as Word, PDF, or plain text.
We just launched version 2.0.0 on Microsoft Store: https://apps.microsoft.com/detail/9nqjkq3l649b
Check out demos: https://ekhos.ai/demos
Benchmarks & hardware requirements: https://ekhos.ai/help-center/hardware-requirements/
Learn more on our website: https://ekhos.ai
What’s next:
Additional on-device AI tools (e.g., summarization) and MacOS support
Please give it a try. I’d love to hear your feedback and answer any of your questions.