Link: https://pivot-to-ai.com/
What Web3 is going great is for Crypto(currency) stuff, Pivot to AI is for AI-related grift.
Anthropic: Could AI be conscious? Pope Leo: lol, no
The Catholic Church has a number of obvious issues. But it’s not hard to do better ethically than an AI vendor.
This year I first used it for post-processing the videos for the German Perl Workshop. Usually
the videos only need some slight cutting or insertion of a simple overlay, which ffmpeg does well enough. This year,
I wanted to blank out some PII from a video, and having an interactive
non-linear editor was convenient here.
The program supports Linux, and it is a monster of an non-linear video editor. The free (as in beer) version can do everything I need, the paid version is cheap(ish) at € 300 (?) for a permanent license. The permanent license has more filters and some AI built-in, but I don't need that. Blackmagic Design make their money from selling hardware, and I also have one ATEM Mini Pro HDMI multiplexer.
The software is complex, but there are lots of Youtube videos showing how to find the relevant buttons. It took me half a day to find out how to overlay a video with black bars censoring some personal identifying information (PII) for our workshop videos. But in the process I also learned that even the free edition of Resolve can do automatic tracking of elements in a shaky-cam video.
For creating the filters, Resolve has the usual filter graph infinite canvas, which also takes some getting used to. But the automatic layout of the graph connections works well enough for me, and I'm not doing anything fancy like camera tracking or inserting dynamic video into other video.
I've used the software on my beefy Windows machine with an RTX 4090 and 128GB RAM, and editing and rendering a video on that machine is a breeze, even on a remote desktop connection from my main Linux desktop. On Debian, I didn't get it to run since it wants a newer version of the Nvidia drivers than Debian stable provides, and I didn't want to play around with mixing Debian stable and the Nvidia repositories.
I took this picture on a photo walk with colleagues, trying out the Canon 100mm macro lens.
Continued from this post, wherein I animate some SVG
After creating the SVG animation I want to export the SVG file as a video file,
preferrably mp4 or webm.
lenra-io/svg-recorder works, but
only with Chrome. I don't know why the img tag with the SVG URL cannot be
fetched as animated bitmap with Firefox,
maybe this is to prevent canvas / browser fingerprinting.
The result is OKish:
The original SVG as comparison:
The default export quality is low, at 1MB for x frames in 20s.
Bumping the video bitrate in the HTML file improves the quality a lot:
videoBitRate: 25000000
There still are some bits that I don't like. Exporting a 1080p or 4K mp4 does not work on my machine in realtime (Linux, Nvidia RTX 2080, Nvidia drivers version 550). The animation stutters and when I try to export 20 seconds, I get less than that in the exported frames. The alternative here seems to manually set the time offset for each frame and then capture it, instead of relying on the capture being fast enough. On the other hand, trying this on Windows, RTX 4090, with the default driver version yields better results, so I'll simply do the final exports from there.
I took this picture at random with my mobile phone. Mostly, I enjoyed the pastel colors and the contrast with the rocks in front.
