User Guide
This is a suite of tools for processing recorded teletext signals as well as tools for processing teletext packet streams.
The software has been tested with bt8x8, CX Cards & the DomesDayDuplciator capture hardware (VHS-Decode), but should work with any VBI capture hardware with nearly any appropriate configuration.
Download & Install Dependencys
Python 3 - sudo apt install python3 & pip - sudo apt install python3-pip
Download the repository via clone
git clone https://github.com/ali1234/vhs-teletext.git
Or via the green code box, local, download zip.
pip3 install -e .[spellcheck,viewer]
For latest versions of Ubuntu like 23.04 the installation is done using
pipx install -e .[spellcheck,viewer]
The decoder can use Nvidia Cuda or OpenCL acceleration to decode the data.
In order to use CUDA accelarated decoding you need to use the Nvidia proprietary driver.
Open a terminal
Update Packages with sudo apt update
Install the Nvidia Driver with sudo apt install nvidia-driver-525
Install the Nvidia Utilitys sudo apt install nvidia-utils-525
Install CUDA Toolkit sudo apt install nvidia-cuda-toolkit nvidia-cuda-toolkit-gcc
Enter into vhs-decode teletext with cd vhs-teletext
Now run this command sudo
pip3 install -e .[CUDA,spellcheck,viewer]
For latest versions of Ubuntu like 23.04 the installation is done using
pipx install -e .[CUDA,spellcheck,viewer]
Without using any sudo.
Now run a deconvolve, you should see a performance increase and no errors about incompatible CUDA hardware
In order to use OpenCL you need to install pyopencl and the appropriate opencl runtime for your card.
Install OpenCL
pip install pyopencl
Using OpenCL Decoding
When running the deconvolve command use -O option.
In order for the teletext output to be rendered correctly in a terminal you need to use a specific font on said window. See Terminal Set Up.
Run each command with --help for a complete list of options.
Then Indexing and breaking it down into pages.
.tbc - Time Base Corrected (Decoded video frames from vhs-decode/ld-decode/cvbs-decode)
.vbi - VBI Area Data
.t42 - Decoded Teletext Packets
Capture VBI via BT8x8, SAA7131, CX88 Based Chips :
teletext record -d /dev/vbi0 > capture.vbi
Capture full video signal RF data via CXADC or DomesDayDuplicator (DdD)
Connect video input & adjust gain level with
./leveladj
Capture via CXADC with
timeout 10s cat /dev/cxadc0 |pv > CX_Card_28msps_8-bit.u8
timeout 10s defines the capture duration of 10 seconds, this can be defined in hours minutes or seconds if a timeout duration is not set it will capture until storage space runs out or stoped manually with Ctrl+C.
Decode Capture with vhs-decode / ld-decode / cvbs-decode etc
./vhs-decode --debug --chroma_trap --pal --threads 8 --cxadc input_capture.u8 output_decode_name
All decodes are in the TBC format with the full 4fsc signal frame for vhs this is just the tape_decode_name.tbc which only has the Y or Lumance data no chorma information.
Deconvolve a conventional VBI area only capture:
teletext deconvolve capture.vbi > stream.t42
Deconvolve a full-frame TBC file:
teletext deconvolve -c ddd-tbc INPUT_tape_decode_name.tbc
While deconvolving some status elements are shown
959392L [13:49, 1156.75L/s, R:44%, M:|▅▁▁ ██|]
- 959392L The amount of VBI lines already processed
- 13:49 The time in minutes and seconds since the deconvolve started
- 1156.75L/s The amount of VBI lines processed within a second. The more the faster the deconvolve is finished. It seems to include non-teletext lines as well, as with fever lines used for teletext the amount increases.
- R:44% Amount of VBI lines rejected. This counts for lines in the VBI, that are not detected as teletext. The less lines are used for teletext the higher the amount is.
- M:|▅▁▁ ██| Each block represents a magazine
View teletext data inside a TBC with:
teletext vbiview -c ddd-tbc INPUT_tape_decode_name.tbc
Examine the headers to find services on the tape:
teletext filter -r 0 stream.t42
Squashed approach to examine the headers to find services on the tape:
Helpfull if the captures are very long and contain a lot of data
teletext scan stream.t42
Split capture into services:
teletext filter --start <N> --stop <N> stream.t42 > stream-1.t42
Display all copies of a page in a stream:
teletext filter stream.t42 -p 100
Squash duplicate subpages, which reduces errors:
teletext squash stream.t42 > output.t42
Generate HTML pages from a stream:
teletext html output/ stream.t42
Interactively view the pages in a t42 stream:
teletext service stream.t42 | teletext interactive
In the interactive viewer you can type page numbers, or '.' for hold.
This is the second rewrite of the original software. The old versions are
still available in the v1 and v2 branches of this repo, or from the
releases page.
- OpenCL Support 30.03.2023
You can see collections of pages recovered with this software at: