Teletext Recovery CUDA Processing Guide

This page will serve as a guide to explain how to improve your Teletext processing abilities by using NVIDIA CUDA.

You must have a compatible NVIDIA graphics card in your Teletext processing system for this to work.

Objective

By the end of this guide, you should be in a position to process your Teletext data ("deconvolve") by utilising the power of your graphics card. This will greatly speed up the process, sometimes up to 10x, meaning more time to process more teletext!

If you haven't already, make sure you have set your system up for capturing Teletext data. This guide will start where that guide left off.

Configuration Process

We will get straight in. I am assuming you have installed your CUDA enabled NVIDIA graphics card into your system, you are getting video output in Linux from it and you have the terminal open and ready.

WARNING: Getting this setup is a royal pain in the butt. Once it is set, I suggest not adjusting your graphics drivers any further. This is why I recommend having a seperate hard disk/SSD used for performing Teletext capture/recovery.
SECOND WARNING: This guide works as of May 2023. Things do often change, I will do my best to keep this as up to date as possible.
THIRD WARNING: This guide is written with the intention you are running Ubuntu 22.04 - it may or may not be the same for other distro's/versions..

• Open a terminal
• Update all of your current packages by typing sudo apt update followed by sudo apt upgrade
• Now run this command sudo apt install nvidia-driver-525
• Followed by this command sudo apt install nvidia-utils-525
• Followed by this command sudo apt install nvidia-cuda-toolkit nvidia-cuda-toolkit-gcc
• Change directory to your vhs-teletext installation
• Now run this command sudo pip3 install -e .[CUDA,spellcheck,viewer]
• Now run a deconvolve, you should see a performance increase and no errors about incompatible CUDA hardware

What Next?

That's it for now, I'll likely start a guide shortly about tips and tricks for "cleaning" up the captures.

Further Reading

If you are a bit lost with all of this, maybe these links may help

• My Teletext recoveries so far
• Teletext Wikipedia article that explains what Teletext is
• ali1234's GitHub repository for the vhs-teletext programme
• The Teletext Archaeologist's website that explains a bit more about how this all works
• The Teletext Archive, as operated by The Teletext Archaeologist himself, a database of recoveries made so far
• The Teletext Archaeologist's twitter profile for various bits of multimedia relating to the format
• ZX Guesser's page about teletext recovery.
• A YouTube video by ZXGuesser performing some advanced recovery techniques with his own toolset
• A YouTube video by The Centre for Computing History of a talk by The Teletext Archaeologist from Block Party 2022 that explains how this all started.
• The vhs-decode project on github is another project that uses RF capture of the video stream, its a little bit more in-depth than what is proposed here, but has reports of being effective.
• The Discord server where the community are ready to help you with any questions or queries you may have