Do you remember this guide? (Amazed it's still available even after 10+ years.)

I know it was quite popular to resize movies in the Quake community, to release versions with different resolutions to suit the viewer's preferences, and how bad their PC was I guess, since not everyone could playback the highest resolution they used back then. Plus x264 required even more processing power, so they even released XviD versions.

What about the CS community? Did you resize anything at all? Do you know of any CS movies that had some sort of resizing? I suppose one reason we didn't see resizes was because of the HUD scaling.

I remember I did some quick tries in 2010, but it felt messy and took a lot of HDD space since I had a bad PC, 1280x1024 monitor and had to use HLAE to force a higher resolution. Took quite a lot of time too. Yesterday I felt like doing a proper try, lol.

(19:03:49) (+Dec) http://i.imgur.com/YfVvouY.png 960x540
(19:04:08) (+Dec) http://i.imgur.com/ttNfV8B.png 1920x1080 resized to 960x540

As you can see, the textures and objects farther away are sharper.

Today however, resizing feels pretty pointless since you could just make your movie in 1080p, upload it to Youtube and let the viewer adjust the resolution. Also one problem could be that even if you wanted to resize from say, 4K down to 1080p, you would have to own a monitor with 4K support, unless there is a way to "hack" your game. (HLAE, but that is only for a couple of games, plus it's broken at the moment.)

Any graphic cards out there that support full-scene anti-aliasing (FSAA)?

Because what you are doing is basically FSAA in combination with additional "traditional" MSAA.

Btw what function (resize filter) did u use for the downsampling?
A Sinc based (i.e. Lanczos) resampling could give "better" results than cubic for example.

Edit: Lanczos based resampling has some disadvantages too and cubic is already pretty good, apart from that it's response depends on the direction of the edges, which is a bit odd.

Yes, I followed the Style guide :P

damn thats actually pretty cool seeing the difference in details at further distances. I wonder if its actually worth it tho since the change is minimal and the action will be too quick to fully appreciate it. though, its all moot now since the status quo of computers today allow 1080p to be played smoothly.

I used to like walle tiffany twisted movie look, so I often resized into wide aspect ratios (capturing at 4:3). Other than that no, HDD was too slow, and HUD scaling problem. I got bored with cs movies once everyone started doing high resolution vids. Personally I thought it looked horrible on 1.6. Not sure if it was just a lower framerate that most people used when capturing in 720/1080 but it looked a lot less smooth than 640/800.

talking about video smoothness:
i've recently started making 1.6 movies at 60fps playback. guys it looks insane. will post when i make smth proper, pretty soon. unfortunately youtube deosnt support it yet.

Not yet. Soon though, they have a few 60fps videos available to show it off.


And yeah high playback rate @ 1.6 looks awes0me. Could anyone make velocity based motionblur for 1.6? Relatively simple openGL stuff as far as I understand.. ~ ripieces come in, over

i have a movie ready 4 upload to youtube when 60fps gets enabled.

my comp doesnt manage the playback @ youtube, old news, but whatever.

what #8 said, ripieces :)

why don't you give it a go, Spunge :D from what I've heard through the grapevine, you're a pretty savvy coder/developer

What makes you think that real velocity based motion blur would be better than what HLAE sampling does?
In fact velocity mapping is inaccurate too for several objects, since it assumes a piece-wise linear movement.

Using a high enough SPS you should get (very) good results with HLAE already.
So I suppose you are just thinking that it would render faster.

Proper velocity based motion blur is not that easy in my humble opinion, since somehow you would need to get the velocity vectors for every fragment into the render.

PS:
If you want motion blur to cross the FPS boundary (cross-frame motion blur) you can use mirv_sample_frame_strength.

PPS:
Also the sampling system in HLAE has been improved and fixed a few times, if you haven't tried it since HLAE 2.2.21.0 (2011-07-12T13:37Z), you should give it a shot:

Ah, it's been a long time since I used HLAE so maybe things are a lot different.

We talked once about grouping the frames (so for example, 20 frames were captured at each REAL frame interval, but all very close to the original time (tiny offsets, not spanning the full gap)), did you ever add something like that? I remember trying the weighting or something similar, but I think that captured all frames at a consistent rate and just weighted the blending of them rather than dropping them altogether if I remember correctly?

And you're right, capture speeds were/are my main concern. And yeah I guess @ proper velocity blur you'd have to track all movement. I was comparing it to just using the modelviewmatrix which I'd imagine would be easy to implement. I didn't consider having to track all fragments.


Also, PMing you about something unrelated, heads up~

Have a look at mirv_sample_exposure in HLAEwiki which implements control over the frame exposure time. It's not exactly what you wanted, but close to that:

This is similar to a Rotary disc shutter where
mirv_sample_exposure = shutter_angle / 360°
Of course HLAE does not simulate a rotating disc, instead the whole frame is closed or opened ideally at once (electronic shutter).

However HLAE causes the engine to render all samples during that frame (because I was afraid making the engine skip might cause problems), it just skips the samples when the shutter is closed (doesn't compute them). Still somewhat fast.

CS:GO update! >:D

I got reminded that Nvidia (not sure about AMD) has support for DSR which allows you to use a resolution higher than your monitor supports natively. So I gave it a shot:

1920x1080
5120x2880 downsampled to 1920x1080

(Imported the .tgas directly to Vegas this time, dunno how Vegas handles downsampling.)

Still, I am not sure if it's worthwhile to downsample? Might as well keep the highest resolution as long as you have space for it. The interesting thing is DSR itself though; being able to render graphics in a higher resolution to improve the image quality, which could be cool to test in older games, since newer games sometimes have a render scale option :P

One downside in CS:GO is that the console and UI will look very small and difficult to read, + the mouse sensitivity is affected.

Though, those who want to record in a higher resolution than their monitor supports can use DSR. P.S. It only works in fullscreen mode :-)

now this is interesting.  I wonder if you could trick the youtube bitrate algorithm with this.

Here's a resizing example on a depth stream dedicated to PANIq:

4K downscaled to 1080p vs 1080p