@PikkuLasse@UnrealEngine That jitter in the profile/side tracking shots is a quirk of take recorder. There's a fix, you use take recorder to record at a slower speed. The extreme end is to Set Engine Time Dilation to 0.1x and set take recorder framerate to 240fps. Only noticeable on side tracking shots.
Vigo the Carpathian! Made this build for Halloween! Vigo watches the closest person and barks his lines at them.
Powered by a @Raspberry_Pi@godotengine and YOLO vision model.
Build video coming soon!
@SlowJoCrow1@geerlingguy In film this may be referred to as ADR of you've heard that. It essentially means re-recording audio for missed statements, unclear voice in a particular sentence, or to correct a mistake, like a product number being wrong.
@___frye I think it's a combination of a few factors. Varying brightness of the different subpixels, varying default orientations of the screen, and possibly viewing distance/effective pixel density. The subpixel shape can have impact on antialiasing and text rendering for instance.
@spacegho_st@geerlingguy I checked after the LTT video and I saw the single units were sold out so it defaulted to the 5 packs. I think that's what people saw. They saw the price only, but didn't look at the selected variant.
@theibradev @0xFryl@angerman Probably whatever underlying functionality used in the Python service that is implemented in C in CPython is more optimised than whatever the newly written Go service does. The DB cost is unchanged, but the overhead of the service execution time is larger.
@TomWatkins1994@0xFryl Yup. This whole situation would have been avoided by simply measuring and seeing:
Time in Code: 10ms
Time waiting for DB request: 500ms.
It makes no sense to touch the service code in this instance because the Maximum Impact you can have is 2%.
@TomWatkins1994@0xFryl Realistically, don't start optimising ANYTHING without first measuring. So many easy wins or trivially discoverable bottlenecks ignored.
@Ian_Fisch@Jonathan_Blow@substrate_dev Worse, how is it a debate that keeps recurring! They both have their place. But when you need a breakpoint, need to traverse the call stack to understand more context, logs don't even come close.
@Sortospino@TrippSmith_com@Perpetualmaniac@ianmiles 100 bucks an organisation that lets ChatGPT code go unreviewed and skipping a staged release process out globally to millions of systems has a massive process failure. Also, never in the history of programming has a memory safety issue reached production before ChatGPT...
@exifrexi@MartianDays@rms80 Yeah I read it in context with another one of his comments in that thread. You are absolutely right though, all too often even prototype stuff that's meant to be replaced ends up shipping.
@exifrexi@MartianDays@rms80 I interpreted Ryans post more about iteration and experimentation without much comment on runtime implications. I think there's huge benefits to iterating fast in BP, even if poorly and then refactoring, optimising and porting as required.
@simonjester2424@TracketPacer Disappointing, I know. I tried coaxing it into treating it as a real sentence it just doubled down on saying it's a meme and it's notable for sounding real but being gibberish.
@simonjester2424@TracketPacer Lol. I asked Gemma2 and it recognized the meme text and basically gave a summary/cherry picking of the know your meme page.