Animation playback on the TI-99/4A

[TheMole]

 

Not quite, but we could try. They probably got a PC handling logon, communication between Netflix and NES, apart from realtime dithering of images before sending the "image" to the NES. And then a 256K cart to do the GUI etc.

 

I don't think they actually did all that... it seems mostly fake to me and I think House of Cards might be the only thing they have converted.

[Tursi]

Before the video, I was thinking about how it might be done and had convinced myself you could easily hack up a PC->NES interface to do the heavy lifting.

 

After watching the video, especially how staged it sounds, I agree with TheMole - I suspect they just faked it. That would explain why there's no dithering and needs 256k, it probably actually is running all on the NES.

[+Schmitzi]

yes, should be a fake, of course, or how did they get that ad on the TI ?

[+OLD CS1]

yes, should be a fake, of course, or how did they get that ad on the TI ?

 

Those advertisements are the reason that I cannot watch YouTube videos. Between Ghostery, FlashBlock, and AdBlockPlus, I cannot get a single YouTube video to actually play. Most of the time I just download them to watch them if I deem the material important enough.

[+Ksarul]

I actually went and did a careful set of experiments to see which plugins were causing me the problems--the only one that mattered was FlashBlock. The others usually won't stop what is going on from the video's standpoint.

[Tursi]

Do the plugins actually stop the overlay ads?

 

For my videos, I always run them without ads for at least a few weeks before I go back and add them in, but I only use the overlay ads (not the ones that play before the video). If ad-blockers can't watch my videos because of the overlay ads, I might turn those off too. It takes about 3 years for me to get a Google check anyway

[+Schmitzi]

Hi, I do not experience any problems, maybe because I use the AvantBrowser.com ?

This is really a good browser, maybe you have a look. It runs parallel to others.

 

It comes in the Standard-version, without an engine, just like a "skin" over the IE,

or in the Ultimate-Version, bringing the actual Mozilla- and the Chrome-engine with in.

(fully parallel to existing installations)

 

In the Ultimate, you can switch between the engines, per register-tab and/or per domain,

and you also can pre-define which sites should run in which engine.

 

There is a ton of (non-disturbing) additional features, like auto-form-fill and i.e. switch flash&Ads... on/off.

The only issue after install is, that the first step should be to go to the ->options-menue,

->Skins and choose ->Windows-Traditional, to get rid off the very first "funfair"-setup

ah, and maybe you should say "NO" once to the first "store passwords in the cloud"

(this is for synch to your other PCs)

Ralf

[Tursi]

Updated my conversion tool suite to fix a few glitches that had been annoying me:

 

-Added proper support for non-widescreen aspect ratios. Everything should now properly scale to fit the screen (no more white bars on the right).
-Less noisy audio playback (removed dither on SOX)
-Added F18A GPU enhanced playback. This is autodetected and (currently) can not be defeated (plan is for the later final version to have a key or config override). It is not yet tested on hardware, any feedback is helpful!
-Fixed timing of cartridge playback to match non-cartridge version
-Updates to cartrepack and both versions of the playback code to account for the above.

http://harmlesslion.com/software/tividconvert

 

The big one is the F18A playback (assuming it works on hardware, but it's just code, it should). This version works by uploading each 768 byte block of video to a fixed, offscreen buffer in VDP. The GPU in the F18A then moves it VERY quickly (I didn't use the DMA functions as that feature isn't stable yet, after 1.6 is well spread maybe).

 

The playback code divides the screen into four horizontal bands, and alternates pattern and color uploads, as a tradeoff between upload speed/framerate and flicker. It works okay most of the time, although extreme color changes (like white on black) can be very distracting. (I tried other tricks to reduce that, like being smarter about changing the colors back and forth, and so far every attempt has looked worse).

 

For the GPU version, after each band, one more byte is written telling the GPU what to do with the block. For the pattern table, it is written to a second offscreen buffer in high VDP memory (above 16k - mostly because I was out of room in the main 16k). After the color table block is uploaded, the GPU then copies the pattern and color tables into their correct place on screen, alternating every byte. Between the massive speed of the GPU, and the ability to directly access VDP RAM, this means only microseconds pass between the pattern and color table updates, and the only way you'll see the flicker is via very bad luck of the scanline happening to coincide with the pattern table write. I'd be surprised if it was more than a few pixels wide, and it would definitely be only a single frame.

 

In emulation, at least, the result is flicker-free, which looks much nicer than I expected it to. I'll do a comparison video in a bit (takes a lot time to upload to YouTube). So I've made it part of the standard code.

 

This is still pretty much still exhibition status, I know, but I know where I'm going with it. It should be usable as an actual playback function soonish.

[Tursi]

When YouTube finishes uploading, the GPU demo video will be here (re-upload of White and Nerdy, I took down the old unsync'd versions):

 

http://youtu.be/3PGCCXRefZM

[sometimes99er]

... This video contains content from SME, who has blocked it in your country on copyright grounds.

[Tursi]

Blocked world-wide this time. Good $%#@ lord. The old video had no restrictions at all. And surprise, SME again. Screw it.

[Tursi]

On the plus side, the a/v conversion is good enough to trip their detectors, neh?

[+OLD CS1]

When I start making videos I will be hosting them on my own website. Screw YouTube and the rabid dogs which police it.

[+Vorticon]

And we'll call it Pirate Video

[+OLD CS1]

And we'll call it Pirate Video

 

Just do not call me "Dread Pirate Roberts" as that was co-opted. I will, however, accept "Happy Harry Hard-on" (for you Pump up the Volume fans out there.)

[+arcadeshopper]

Ot bad apple on vectrex

http://vectorgaming.proboards.com/thread/1309/bad-apple-on-vectres

Edited March 23, 2015 by arcadeshopper

[Tursi]

I was asked for a copy of the cart image for my videos, so while they're online, grab them here:

 

http://harmlesslion.com/temp/vids/

[+Ksarul]

Thanks, Tursi!

[+arcadeshopper]

I was asked for a copy of the cart image for my videos, so while they're online, grab them here:

 

http://harmlesslion.com/temp/vids/

 

Looks like the target for the next cart is 16mb

[Tursi]

If anyone is using my tools linked above, go download them again. My fix for long videos didn't work, and it bit me again. It's fixed properly now.

[Tursi]

Since I just linked the Coleco guys here, I should link to the YouTube vids that didn't get taken down. (In addition, the temp/vids folder linked above not only still exists, but I add to it from time to time ).

 

https://www.youtube.com/playlist?list=PLr4KHMd3l8gh0q9bYJVyDh7T2UerCtFrs

[Kiwi]

I decided to play around with tividconvert. The burger time clip I made, unfortunately, didn't have sounds. I tried it anyway, indeed didn't work. I'm so stubborn... Yes I did read the instruction before hand . Anyway, I had a clip of sparrows taking off in slow motion to use as a reference for my "A Sparrow Goes Flapping" for the Colecovision. I converted the video to TI format, since burgertime clip took some time to convert over. So the clip was 52 seconds long.

 

youtube video:

Bin: SLOWBIRD.bin

[JamesD]

Since I just linked the Coleco guys here, I should link to the YouTube vids that didn't get taken down. (In addition, the temp/vids folder linked above not only still exists, but I add to it from time to time ).

 

https://www.youtube.com/playlist?list=PLr4KHMd3l8gh0q9bYJVyDh7T2UerCtFrs

Nicely done.

Perhaps a bit more processing for the audio could reduce the noise?

Something similar to the vertebi encoding some guys are using on the MSX perhaps?

 

The description of the video links to the code.

Edited June 30, 2017 by JamesD

[Tursi]

Sure, let me know when you get it working!

[Tursi]

Anyway, flipness aside, I looked at this when I did the audio conversion. The MSX chip can output a true flat line - the TI chip can not. Even the SMS variant can output a flat line.

 

The inability of the TI chip to do so makes any system that attempts to stack the three channels for more accuracy doomed to failure. I did a fair bit of study on this idea and even created a converter some years ago - well before trying the video conversion. Because each volume step is on a logarithmic curve, in theory stacking the volumes should allow finer resolution. I plotted it and estimated 10-12 bits worth of usable audio range, but I was completely baffled when the output waveform was completely random looking.

 

Ultimately, I worked out what the problem was. The three tone channels continuously switch between positive and negative output at a rate based on the frequency, and a level based on the volume. We are playing back audio here (and in most TI playback programs) by setting the highest frequency we can and manipulating the volume. Right off the bat, this causes interference in the waveform, but the resulting audio can be understood.

 

The problem with stacking the audio is that we can not control the phase of the tone channels. There's no way to force a synchronized reset, and because the chip starts in a random state, the channels are always out of phase. Even if you change the frequency, it doesn't take until the current cycle completes. This means that some of the channels will add to the final waveform and some will subtract from it, with the result being nothing like you wanted. Worse, you can't even tell what you're going to get.