Hacker News Reader

Comment by eli | Hacker News Reader

That's funny -- I just did the same thing! We both must have read that same earlier post about using Visionect to display the NYT and ran out and ordered one.

I didn't get a chance to write it up yet but here's my software https://github.com/elidickinson/newsvision

I skipped the HTML renderer and used the python library to push image data right to the display's frame buffer. Seems simpler if you're doing the hard work of turning it into an image yourself anyway. I also grabbed the full list of newspapers and used imagemagick to annotate each image with the paper name and location. It's a little hacky, but seems to work.

It's admittedly an expensive display, but it really is very well made piece of hardware.

View on HN

The earlier post, if I’m not mistaken: https://news.ycombinator.com/item?id=25063726

Yes exactly. There was also this one prior where the author built the panel/frame themselves: https://onezero.medium.com/the-morning-paper-revisited-35b40...

Is this the model you use? https://www.visionect.com/product/place-and-play-32/

Yes. The hardware looks great.

Some other ideas: overlay additional data or insert other screens in the rotation like a local weather map and bus times. Add another mode to display portraits or fine art. I also want to try different dithering and scaling techniques to see if I can get the text crisper.

I also think it’d be neat to make an alexa skill for it so you could issue voice commands. Maybe have it show the song and artist data for what’s playing on Spotify.

I thought about this. It would be a throwback to Pointcast (Circa 1990) for anyone who remembers that. They did the same thing... news, weather, stocks on a rotating schedule.

> the python library to push image data right to the display's frame buffer

How are you doing this? It looks like you’re pushing it to their cloud service (or your local docker copy, if configured as such)

I’m using their cloud hosting now (I think it’s free for the first month) but it would work the same once I’m self hosting the Visionect server. Just change the url to push to localhost and run both containers on the same box. Should be low overhead with the HTML rendering disabled. Could run it on a cheap cloud server or I’ll probably use an old laptop on the local network.

I didn’t understand going through HTML to render it with a headless browser instead of rendering the PDF directly with Ghostscript or something like that.

I used poppler-utils but the OP apparently used imagemagick on PDFs directly which I didn’t know was possible.

(To be fair the device does HTML rendering by default and the API for pushing images isn’t super obvious)

2 reasons I'd use HTML

1. I'm taking untrusted PDFs and running them on my machine. Using something like say puppeteer that PDF renderer runs in a sandbox. Ghostscript, not so much. PDFs can have code in them

2. Using HTML via puppeteer lets me render tons of other things I might want to render than just PDFs. Plenty if beautiful algorithmic art out there that runs in a browser.

Thanks for posting. I didn't see a way to post directly to the display. I'll have to dig into.

Reply on news.ycombinator.com