Article proposal: Whisper running locally without dedicated GPUs

Team Workflows Fedora Magazine
,
1

Article Summary: How to build whisper.cpp locally with an OpenVINO backend on machines without a dedicated GPU, and use it for private, offline voice dictation.
Article Description: This article would walk Fedora users through setting up whisper.cpp with OpenVINO acceleration on Intel hardware. It would cover what whisper.cpp and OpenVINO are, how to identify whether your CPU’s instruction set and integrated graphics can benefit from OpenVINO, how to compile whisper.cpp with the OpenVINO backend in an isolated Python environment, how to optimize performance on systems with Intel Iris Xe graphics, and how to run inference on a converted model. The focus would be on privacy-conscious users who want fully local, offline speech transcription without requiring an NVIDIA GPU.

I have read and understand the Ai-Assisted Contributions Policy

2

This sounds like good content for Fedora Magazine. +1.

3

+1 from me.

Please use comments on this topic if you have questions or need to communicate with the editors about anything.

The overall work flow is described at this link as well as other helpful information.

This site might contain some items of interest

The articles are written using the WordPress instance for the Fedora Magazine.
From the menu in the left column of that page, select
*Posts>Add New Posts
to open the new article to edit and get started.

When you have it ready to review in the Fedora Magazine WordPress site, please leave a comment in this topic, with a preview link, and we will start the review process.

Thanks for volunteering to write for the Fedora Magazine!