[X] My Assistant

Those talks make me realize once again I really understand nothing about encoding.
Last time I tried to do it, result was ugly, despite doing the best I could: using the slowest mode.
I also have no idea if there's anything that could be done to fix videos where there isn't a keyframe at a transition between 2 scenes, thus potentially causing bugs or crashes if a chapter is put at this position (though I guess what cause that get fixed a little bit over time)

edit: following below answer, I bolded something (IMG:[invalid] style_emoticons/default/laugh.gif)

This post has been edited by uareader: May 19 2021, 23:18

Those talks make me realize once again I really understand nothing about encoding.
Last time I tried to do it, result was ugly, despite doing the best I could: using the slowest mode.
I also have no idea if there's anything that could be done to fix videos where there isn't a keyframe at a transition between 2 scenes, thus potentially causing bugs or crashes if a chapter is put at this position (though I guess what cause that get fixed a little bit over time)

i haven't seen missing keyframes cause crashes, just ugly video.

also set your quantizer minimum/maximums. That'll have a lot more impact than using the slowest mode on output quality. the slow modes basically just give the same quality at a smaller size, but they yield diminishing returns.

Just fixed a bug in an emulator and got it accepted upstream

Feels good

I know political nonsense is frowned upon, but I was reading a snopes article that assumes that most people lack any technical knowledge about hard drives. Please correct me if I am wrong.

The tech team doing the Maricopa county audit sent a request saying that it appeared the databases had been deleted and requested copies from the backups.

The election officials said that the files were not deleted, the tech team lacked the expertise to reconstruct the file system to access the directories and that their backup was perfectly fine.

Snopes says the files were not deleted. This technically true if the hard drives were formatted, as only the file system is deleted not the files themselves with a fair bit of effort can be usually fairly accurately reconstructed....

Hmmmm The full technical response claims that the tech team did not reconstruct the raid array properly which corrupted some data and also that several databases were not removed, they were moved to backup storage to reset the voting systems for the march Jurisdictional election.

[www.snopes.com] https://www.snopes.com/uploads/2021/05/1-Te...ponse-Final.pdf

Thats probably why the tech team was asking where the backups were...

Normally when you do a project like this wouldn't you have liason with the original system administrators though I guess because its a legal process they just hand everything over without explanation.


Hmmm whenever I've had an array fail or reconstruction fail usually the impact is bigger than a few missing / corrupt files. I guess they were the biggest files?

Yeah, last time I had an array fail I lost loads of stuff.

Thought:
I wish more sites tried to stick to 80 columns in their source HTML where possible. It's always a pleasant surprise to see.

I know it's not a good idea in lots of places, but for large paragraphs of text it's quite nice and makes a page a lot easier to read the source of.

I guess I can't really expect that in this age of frameworks, Sass, and auto-generated content

I was about to say, that it depends which platform or editor they are using these days.

I opened up a sitemap the other day and crashed my 30 tab browser.

I wish more sites tried to stick to 80 columns in their source HTML where possible. It's always a pleasant surprise to see.

+1 +10^54

I opened up a sitemap the other day and crashed my 30 tab browser.

Woohoo!!!
...
Ah, sorry for your browser, but it made me think of a ps2 emulator site, I went to check it, and miracle, after long years (and many messages sent about it), they finally have something else than an error 404 (or whatever error was there) behind their "sitemap" link (IMG:[invalid] style_emoticons/default/rolleyes.gif)

With h.264 at least, there's a tuning preset for animation with all the flat surfaces.
Did you try using a tuning preset like that? I think there's one for HEVC (h.265) as well.

Tested the animation preset first on an episode of Magical Sempai (due to its short runtime of 11 minutes per episode). Encoding at CRF 20 Slow, final video bitrate was reduced to 1.5mbps.

According to x265's doc, the animation preset reduced the values of AQ-Strength & Psy-rd. Found some info on what those options do after a doing some search:

aq-strength is the strength of the adaptive quantization offsets. Default is 1 (no offset). Higher = tendency to spend more bits on flatter areas, vice versa. Setting <1 in crf mode decreases overall file bitrate and reduce spending bitrate on plain areas (but potentially introduce blocking/banding in higher crf).

psy-rd will add an extra cost to reconstructed blocks which do not match the visual energy of the source block. In laymen’s terms, it throws in extra bits to blocks in a frame that are more complex. Higher strength = favor energy over blur & more aggressively ignore rate distortion. Too high will introduce visual artifacts and increase bitrate & quantization drastically.

Another trade-off like SAO.

Also, how is the framerate? I've not watched any TTG, but anime has a pretty low framerate typically, so if TTG is more fluid that is definitely part of what happened.

Typical constant framerate of 23.9fps. There are action and dark scenes and but the FX are nothing fancy.

Typical constant framerate of 23.9fps. There are action and dark scenes and but the FX are nothing fancy.

I mean, even at ~24fps, usually if you frame advance you'll see that only every couple frames actually have any changes on-screen with a lot of anime (which is often done at far lower rates).
Video encoders can save a ton of space if basically nothing happens during a frame because the frame delta is small. But if TTG isn't animated "on thirds" or whatever like anime often is, you're not getting that space saving.

Looked up how much a Roland MT-32 goes for on ebay these days, and I'm shocked/depressed

I already have one, but I feel bad for the people who don't now. I mean, come on, this is a late 80's MIDI synth. Who cares if King's Quest used it.

This post has been edited by dragontamer8740: May 21 2021, 08:44

I installed a virtual Windows XP, and I was looking for a way to activate it.
I was so glad I found someone sharing the link [microsoft.gointeract.io] https://microsoft.gointeract.io/interact/in...mp;token=0Yr8Nd to basically using the phone process without actually using a phone.
This was to activate a Windows XP for an always offline virtual machine though. There are so many parameters in that link I would not be surprised if some information could be intercepted and used by the one providing the link for an online system (though maybe I'm being a bit paranoid).
Still, much safer than trying to download and use a crack or something.

I installed a virtual Windows XP, and I was looking for a way to activate it.
I was so glad I found someone sharing the link [microsoft.gointeract.io] https://microsoft.gointeract.io/interact/in...mp;token=0Yr8Nd to basically using the phone process without actually using a phone.
This was to activate a Windows XP for an always offline virtual machine though. There are so many parameters in that link I would not be surprised if some information could be intercepted and used by the one providing the link for an online system (though maybe I'm being a bit paranoid).
Still, much safer than trying to download and use a crack or something.

Just use a volume licensing key (VLK), lol. VLK's don't need online activation for XP.
I have a disc that was purchased from a university for use by its staff back in the day. Warning: it has no service packs installed; you'll have to install either SP1a or SP2 and then SP3 if you do this (or slipstream the disc).
~489 MiB, disc label WXPVOL_EN.
MD5 checksum 687dd0dcff628bbbc237ce534e98e992
SHA-1: 26a053e210e7b1f1cd42bb323d20be95c0d2bc4a
SHA-256: 9cc693df90a448de473ec33853a9e61ced61436f9f5d1e2499f7ecca2b064955

BTW, i found what looks to be the same thing on archive.org. I ripped my own CD, though.

Alternatively i'm sure you can find an SP3 equivalent Volume licensing image.

This post has been edited by dragontamer8740: May 21 2021, 17:16

Just use a volume licensing key (VLK), lol. VLK's don't need online activation for XP.

I didn't know those existed for Win XP, I just went the "normal way" because I still had some XP versions from formations with MSDN Alliance stuff, so I had legit keys/CD images I could use.
I wonder if some of the things I still have could be using those VLK...

But for the record, in case it was unclear in my previous post, I did manage to activate that WinXP.
Now the challenge will be to see how much work it is to get things from my old laptop working on it (I said it in the past, but I miss the MS-DOS era where you copied a folder, and the whole thing in it worked wherever you put it).

edit: urg, after going that far, the choppy sound from loading the VM through RDP (only way to get sound with Hyper-V) make it quite pointless.
Seems like I will have to get VirtualBox or something after all, even though I really wanted to avoid more unnecessary stuff (IMG:[invalid] style_emoticons/default/sad.gif)

edit: ended up removing Hyper V to get the old version of Virtual Box that support 3D acceleration. It was a pain, and I ended up installing it on my SSD accidentally and don't want to do another removal/reinstall (did lots of those).
But seems performances can be here, though may not be useful everywhere from 1st experience.

This post has been edited by uareader: May 23 2021, 20:03

I think variable frame rate was better for filesize than constant, this allows you to crank up the bitrate and with VP9/H265 it looks solid and smooth.

This post has been edited by Pillowgirl: May 22 2021, 00:23

"Hey, guys, where should we put the headphone jack?"
"IDK, what about as close to the cardbus slot as humanly possible?"
--Apple engineers, 2004

(IMG:[i.postimg.cc] https://i.postimg.cc/zGTdcPm2/IMG-20210521-223317-800.jpg)

Meanwhile my thinkpad has them on the complete opposite side of the machine.

My Dell Latitude (a D610, also from 2004) has an even worse design than apple's, though, since it's noisy even when there's no expansions added and the audio circuitry appears to be on the main board (it's on a daughterboard for isolation on the powerbook and my thinkpad, but just below the DDR RAM slots on the latitude).

edit: urg, after going that far, the choppy sound from loading the VM through RDP (only way to get sound with Hyper-V) make it quite pointless.
Seems like I will have to get VirtualBox or something after all, even though I really wanted to avoid more unnecessary stuff (IMG:[invalid] style_emoticons/default/sad.gif)

Could try QEMU with KVM. Works pretty well.
That's what my XP VM uses. But I'm on a linux host.

Virtualbox should also work, of course. But I keep a Pentium M laptop with XP on it to avoid this sort of stuff, though. The VM is just for fun.

Cross-compiling a new kernel for my Powerbook for the first time in about six months.

This post has been edited by dragontamer8740: May 22 2021, 08:21

Wow, mcomix runs a lot slower under Pypy than it does in CPython.

I guess it's making heavy use of C extensions.

If it weren't for the fact that I hate GTK+ programming so much I'd probably try to redo its basic features in C.

This post has been edited by dragontamer8740: May 22 2021, 19:17

YouTube found a way to get around adblockers for the companies that whored out enough money for their ads.

"Hey, guys, where should we put the headphone jack?"
"IDK, what about as close to the cardbus slot as humanly possible?"
--Apple engineers, 2004

Louis Rossmann after dealing with generations of Apple products he would say that their style is reusing as much of the previous generation's design as possible even if its broken or faulty.

Yeah, I know that much.
But the aluminum and titanium books were still pretty new, so not earlier than around 2002, I think.

I'm trying compiling mcomix and its dependencies with Cython now, just to see how much of a free speed boost that might get me.

Working alright so far; I might have to do some of the explicit typing stuff to make it really improve, though. Took me a while to figure out how I needed to do it, though, and the compilation process is quite slow (not too surprising since it's machine-generated code... reminds me a lot of how Adobe AIR worked on iOS, really, where it has to ahead-of-time compile all of its actionscript to sidestep Apple's restriction on interpreted languages. It took ages to do it, even on decent hardware. And I'm doing this on a Powerbook G4.

Trying to cythonize PIL (pillow, a python imaging library) now.

My only prior experience with Cython was decompiling a Cython program that someone had made and released only as a windows binary that wouldn't work in wine, changing the text encoding to UTF-8, and releasing the decompiled app just to spite them. That was several years ago.

They'd released the tool on a forum and hadn't posted in several years, so I doubt they cared much.

Update: can't get Cython PIL working and I'm too lazy/tired at this point to figure out why; I suspect that it's conflicting with an already-installed version of it.

But even without using that (just using non-compiled PIL), compiled mcomix does seem a bit faster. I need to do proper measurements to be sure, though, since placebo is a thing. But I think it might be paging through images/resizing them a little bit faster, regardless. I don't know if startup time is actually any better, though.

--

Just ran iperf on my main trifecta of machines (thinkpad x201 tablet, powerbook g4, and my desktop).

• Note that I am in a congested area on both frequency bands.
• I rounded all these speeds to the nearest increment of 5.
• The powerbook is (as expected) the slower of the two laptops when running over wi-fi (802.11n, ~5GHz) by a sizeable margin.
• Using the Atheros card, the powerbook hits about 80 Mbps (megabits) connecting to my desktop, which is wired to my router's switch via cat6.
• When using the powerbook's integrated "AirPort™" 802.11g card (BCM4306) instead of my ar5416 cardbus adapter, it only hits around 20 Mbps (again, megabits). So my $10 new old stock wlan card approximately quadrupled my wireless throughput. At the cost of headphone noise.
• My Thinkpad X201 tablet has also had its WLAN card replaced (Intel 6200AGN → Atheros AR9280 on AR5BHB92 card), but it's an internal replacement, and I have absolutely no idea where I put its old Intel card when i did this a few years ago, so I can't do a proper "before and after test."
• The thinkpad's getting ~130 Mbps (802.11n, 5GHz) connecting to the same desktop.

I'm still quite happy that 802.11n was drafted early enough for manufacturers to actually make cardbus cards for it. For whatever reason, it's still a surprise to me.

Also I updated its kernel to 5.13 rc2, which I built from torvalds' git master.

This post has been edited by dragontamer8740: May 23 2021, 20:40