So, I’m builduig a case for my actual TV ambientlight setup. Because right now it’s a bit messy. The setup is one rapsberry py 4 model B and one HDMI capture. The rapsberry run with Hyperion.
I just noticed that I forgot to put some ventilation to cool it down.
I added more way to export your ASCII art such as PNG, JPG, GIF and MP4. I also tried to optimize the design to be more user-friendly.
Now I think I have finished this site (maybe if someone propose me a big feature I will come back to my decision). Time to ship the final version. Thanks for all your feedback it was very useful.
I thought it would be easier but I spent 2 hours of finding a good way to convert Gif to video format (such as MP4). I started with a librarie but I had an error, so I switched for FFmpeg and this has led to a lot of bugs. I coorected them and added a better preview of what your video would like.
Now everything works well so try it by converting your best meme.
After hours of development I added the video to ASCII convertor. If you tested it you will see a notification informing you about estimated pourcentage and it can take time so don’t try big video (a big video it’s already a 30 secondes one) or decrease other options.
How it’s work. Basically, it cut video into frame, uses the image to ASCII converter and then recreates a video.
It was such a really challenging idea to implement. My first idea was to call a google API, make the search and take the first image proposed by google. But the API I was looking for was disable for new people. So I try an API with AI generation but it wasn’t free. After that I looked for the reserp API a google scraper. But I needed to stock a private API key on a server and after one hour of trying, testing and fixing bugs. I said to myself that the best way was maybe the wikimedia API, it’s maybe less accurate than other API but at least it was free and no API_KEY needed. So I started devloped this idea and now it’s online. Don’t hesitate to test it and give me feedback or other features you want to see.
Firstly, I added the invert button and the drag and drop zone, it was quite simple. After that, I added the progress notification and tried to optimize the algorithm. It was not simple, I changed it to read only one time the image and not read for brightness and then another time for the drawing… I, also for the black and white version, drew not each character one by one but only after a line. It speed up the calculation time a lot.
I add a section in the README with every features request. I think the first two are excelent ideas, for the third one I have no idea how to make it and then the last one can be a good feature. If you have any ideas I’m taker and if you know how to optimize the algorithm let me know.