Welcome to Fractal Forums

Bump. Video is uploading and should be live in a few hours...

https://www.youtube.com/watch?v=8RF9uiS9VHc

Download high quality (1080p60 H.264 CFR=24)
https://mega.nz/#!L0s0GA4Z!F0yCijoIXiyrdxJiSxE15blfXDRWqYSXBi3XB1gbs_E (https://mega.nz/#!L0s0GA4Z!F0yCijoIXiyrdxJiSxE15blfXDRWqYSXBi3XB1gbs_E)

EDIT: Upload failed. Youtube not support HEVC. And I'm not convinced it is ready for prime time yet. I also had to install the K-Lite codec pack on my 8.1 laptop to be able to play it. Also there were some noticeable compression artifacts in the output so I'm not completely satisfied with the results...

EDIT2: Reencode complete and uploading. H264 (CRF=24, slower) actually resulted in a smaller file with better temporal quality than the HEVC test. Should be live in a few hours. For real this time.

= = = = = = = = = =

I've had this on the back-burner since August 27th. I am doing a zoom movie that passes through the minibrot centered at my contest winning entry, Dance of the Fire Dragon II:
https://www.fractalforums.com/index.php?action=gallery;sa=view;id=19014

It is slow go currently but frames are starting to render faster, between 3-8 hours a piece now. Over the past nearly one month, I've had about roughly 72 hours downtime do to CPU crash, so I have reduced the overclock slightly. :fiery: The CPU is running 4.0Ghz now instead of 4.2, about a 5% reduction in speed, and a whopping 30 watts reduction in power draw at the outlet from around 325w to 295.

Anyway here is a preview render, currently at e241 counting backwards from e257. I've got a long way to go yet:
(http://orig12.deviantart.net/07eb/f/2016/269/2/3/zoom_preview_by_stardust4ever-daizq50.png)
Link to sta.sh page: http://sta.sh/01z3u53o1o96

I will keep you guys posted as details emerge. Not sure if the final movie will be 720p60 or 1080p60. If the image is too grainy or the rectangle frame boarders are visible, I'll have to render the output at 720p.

October 1st BUMP. Moving along. Frames are rendering much faster now, average ~2 hours a piece!
:dink: :beer: :beer:

The symmetry has now gone from eight-fold to four fold...
(http://orig07.deviantart.net/735b/f/2016/275/e/5/dragon_zoom_preview_ii_by_stardust4ever-dajo6yb.png)
Link to sta.sh page: http://sta.sh/01kftlt5pcfi

One very obvious change is the star formation is now much smaller in comparison to the halo. With the first 8-fold image, the bottom formation was twice the size of the top formation. With the second 4-fold image, the bottom formation is a whopping sixteen times larger than the top formation. If you look closely, you can see the tiny four-fold star pattern swimming in a sea of red.

Money shot!  O0

(http://orig08.deviantart.net/fbaf/f/2016/277/5/1/firedragon_preview_by_stardust4ever-dajvrnn.png)
http://sta.sh/01h82ggr0a8i

Zoom progress is really picking up speed now. Believe it when they say the last 25% of the movie takes like 90% of the render time.....

That's why I render my zoom movies in reverse!

That way they constantly pick up speed instead of slow down.  Does wonders for the psychology of waiting   :music:

Very nice pics, by the by   O0

Yes thank you. The way things were going at the beginning with iterations in excess of 100,000,000, I was hoping the zoom would be finished by Christmas!  :tease:

Not an extremely deep zoom (minibrot around ~e257) , but the extremely high iteration depth made it take much longer than other deeper zoom locations.

Halfway there. It's now at the final fork in the road. Over a month of render time so far, but at the rate the frames are now being generated, it wouldn't surprise me if I woke up tomorrow (or at least before nightfall) to find it finished!  ;D
(http://orig07.deviantart.net/4919/f/2016/278/7/2/last_fork_50_percent_by_stardust4ever-dak1tzv.png)
http://sta.sh/01uxv4ycx2yc
 :cantor_dance:

Prolly the last update before the render completes. It's currently rendering the close flyby of a Satallite of which I zoomed into the Elephant Valley. I'm only posting a thumbnail of KF because it is dog slow right now.

What CPU are you using for this render?

AMD Bulldozer 8 cores (FX 8150, clocked at 4.0Ghz)

Still stuck in the "Armpit" of the parent Mandelbrot. Here's a screenshot I made yestereday. All that black area is taking forever, over 5 hours per frame. CPU still burning hot... :fiery:

Even at 8 million iterations, it shouldn't be this slow, considering I'm at ~e005 currently and float point should be more than adequate at this level.

Frames are 100% done; working on rendering AVI frames w/ color cycling. This should be an epic zoom movie when I'm done with it... O0

I've missed this!
This looks indeed very promising to be an epic movie!  :joy:

KF, is there some reason frames were taking ~6 hours to render at floating point precision levels? Since my seed was -.75+.001i, the seahorse valley area was mostly black pixels. Maximum iteration depth at the end was at 8 million but this appears to be software controlled. A simple periodicity test to check for repeated values (the main cartoid has a period of 1, and the bulb 2) would drastically speed up rendering of the black area. The guessing algorithm is dormant until 25%.

The only reason is to avoid having too low iterations close to the mandelbrot/minibrots, so it is a trade-off and hard to adjust by the program for each case.
One can stop the render, adjust the maximum iteration count, and then start the render again at that position, since the kfb files are browsed in the destination folder and file number is resumed.

And I agree, when zooming near the edge of the main Mandelbrot or a mini, the detail needs to be defined. This is less critical IMO at the end of a zoom sequence when all detail turns into ugly gray bands.

The color cycling on this zoom movie is interesting. Starting with a seed value of -.75+.001i, deeper in the ravine of the Seahorse valley than most will venture to go, the iteration depth at this point is Pi/2 * 1000 or about 1570. By zooming into the tip of the nearby bulb, any semblance of spirals is lost and the blobs show up as concentric onion like circles with a sharp increase of exactly 1571 iterations between each layer. Moving into the tip of the blobs creates smaller blobs. I zoom some distance into the tip before selecting the centroid of one of the blobs. Period goes from 2, to 4, to 8, and so on until a Minibrot is visible at the center. The cascading chains form columns of about 23 bulbs between each of the onion-like layers. I then zoom into the Elephant valley of this minibrot to form the complex Julia formations which will be further manipulated.

The onion layers are everpresent throughout the entire zoom sequence with precisely 1571 iterations difference between each layer. This causes iteration bands to behave in unexpected ways. Setting the iterations to a multiple of this value create patterns where each layer offers repeated color patterns. Setting the iterations to 1571 or to a factor of this value result in each band having the same appearance. Factors with higher numerators result in higher contrast with every band. Initially I set hue value in the infinite waves pallet to be exactly 1571. In practice, the result of this perfect match is that the color of the layers remains mostly unchanged, resulting in a boring zoom movie. By setting the value to a near miss, 1570, the colors pop out and can be seen to change slowly, creating vibrant contrasting colors when dense spirals are approached. The farther away from 1571 the value, the faster it appears to change in the zoom movie. Furthermore cycling the offset of the waves during playback results in a different rate of change for the colors overall, compared to straight zooming. As a result of this, the coloring of the entire image morphs simultaneously, rather than color cycling which appears to zoom in or out.

To preserve uniform texture throughout, saturation and luminosity are adjusted to be near miss factors of 1571. The more deviation a multiple of the saturation or luminosity has above or below this value, the more rapidly the values appear to change as the zoom movie progresses. For the initial zoom sequence, I had saturation set to a near miss of 1/3 and luminosity set to a near miss of 1/10.

actual period of layers: 1571

H=1570
S=523 (523*3=1569)
B=158 (158*10=1580)

Unfortunately, there is apparently not enough drift of the saturation value in the current zoom frames, and too much drift of the brightness. With a frame rate of 60Hz and 60 movie frames per zoom level, and incrementing the offset one iteration per movie frame, the cycling of brightness is too frequent, resulting in a pulsar effect where the entire frame goes light-dark-light-dark. I plan on changing the saturation to 522 (522*3=1566) so it doesn't align as well with hue, difference of 4 instead of 1, resulting in greater variety of colors. I plan on changing the brightness from 158 up to 315 (315*5=1575), resulting in a lower offset compared to H=1570 and exact periodicity of 1571. This should cut down on the strobing effect of light-dark-light-dark but may also decrease overall contrast in the columns. If contrast turns out to be too low, I can fix this by incrementing the 315 value to 316 or 317.

Sorry for the long explanation.

TL:DR - I have put a lot of thought into the color cycling of this zoom movie, and color patterns created by micro adjustments of the offset will produce color variation completely independent of the changes presented in a static zoom sequence. In layman's terms, the coloring should result in a highly psychedelic viewing experience!
 :spiral:

:thumbsup1:

It sounds like it is going to be an interesting zooming video.

I have added buttons for finding the closest prime-numbers for the waves in order to avoid them to align.

The guessing of pixels in kf is not so smart, it starts by calculating every fourth pixel and then compare nearby pixels for the rest 75%. I too little knowledge in the other methods that I know exist for pixel guessing

Yeah I'm aware of the prime thing, but the values don't need to be prime to be relative primes. Thing is, close but not exact alignment is desireable in this case due to the 1571 period of the onion like layers. All three values should be near miss factors. Having a smallish LCF like 2 or 5 is insignificant. What's more important is the rate of deviation between each layer. So if the actual period between layers is 1571, then the coloring algorithm between spirals becomes the difference to this value. Using my next set of experimental values,

Hue=1571
Saturation=522 (*3=1566)
Brightness=315 (*5=1575) or 316 (*5=1580)

The onion layer increment becomes:

Hue = -1
Saturation = -5
Brightness = +4 or +9

Color cycling each frame on top of this will have the effect of retrograde cycling especially noticeable with hue. The pallet itself will appear to change independently of iteration depth. Otherwise if I didn't take period into account, each layer wuld vary wildly from the next and the entire zoom sequence would look like fruit salad.

Try tweaking your periods. You can vastly change the color patterns in a Seahorse or Elephant valleys, or even make a spiral go chameleon like and nearly disappear!

= = = = = = = = =

My general strategy for "normal" zoom sequences is to forgo primes in "normal" zoom movies for the Fibbonacci sequence. The series diverges on the golden ratio between consecutive terms and it is proven that any three consecutive terms (besides 1) in the sequence are relatively prime.

1-1-2-3-5-8-13-21-34-55-89-144...

Brightness is the low value, saturation is the middle, and hue the highest.

Funny thing. The golden ratio minus one is it's reciprocal. Plus one is it's square.

BUMP. So I'm redoing the animation with the following color cycling:

H=1570
S=522 (S*3=1566)
B=225 (B*7=1575)

1920x1080p60
60 AVI frames per zoom frame.
60 AVI frames per second.
Cycle colors +1 per AVI frame.

The first 3 minute AVI segment (10800 frames) using the new color cycling is finished and I'm liking the color balance. The bulbs almost tend to melt in and out as the luminosity changes. Here's a screengrab  of the approaching minibrot flyby from earlier this afternoon.

Can't wait until it's completed! :)

The PC finished rendering the AVI frames this morning and I'm joining them in AVIDemux. I have to use zero latency option when joining AVI segments otherwise it drops a few frames at the end. I'm using the advanced H.265 codec for the first time and reencoding to MP4 container. 1080p60, 14 minutes and something. I think we'll see tremendous gains in quality versus file size with H265 due to the advanced motion detection. I'm using the default settings except I changed motion detection to star search. Only thing with H265 is set top boxes and mobile devices probably don't support it yet. :tease:

Youtube will probably ruin it again so I'll be posting a Mega link as usual.

BUMP. See first post. ;D ;D ;D

Going to bed; got busy day ahead of me tomorrow. Later... :sleep:

EDIT: Upload failed. Youtube not support HEVC. And I'm not convinced it is ready for prime time yet. I also had to install the K-Lite codec pack on my 8.1 laptop to be able to play it. Also there were some noticeable compression artifacts in the output so I'm not completely satisfied with the results.

EDIT2: Reencode complete and currently uploading. H264 (CRF=24, slower) actually resulted in a smaller file with better temporal quality than the HEVC test. Should be live in a few hours. For real this time. :tease:

Aaand video is live 4 real this time! ;D
https://www.youtube.com/watch?v=8RF9uiS9VHc

Epic!


 :wow:


 :cantor_dance:

(http://st.deviantart.net/emoticons/w/worships.gif) Epic indeed!

Thank you. Mega link added:
https://mega.nz/#!L0s0GA4Z!F0yCijoIXiyrdxJiSxE15blfXDRWqYSXBi3XB1gbs_E (https://mega.nz/#!L0s0GA4Z!F0yCijoIXiyrdxJiSxE15blfXDRWqYSXBi3XB1gbs_E)