Welcome to our community

Be a part of something great, join today!

Log in
  • Hey all, just changed over the backend after 15 years I figured time to give it a bit of an update, its probably gonna be a bit weird for most of you and i am sure there is a few bugs to work out but it should kinda work the same as before... hopefully :)

M-X Reel

Even more amazing than this reel is the fact that I can soon own one (mx) for 4750
double plus wow !
 
Graeme Nattress said:
It's not on the original 4k for sure.

Graeme
Click to expand...

Goes to show that the whole workflow is important.

Time for a rerender...

Very cool reel. Want. Scarlet. Build one already! Any indication how the lowlight performance of Scarlet 2/3 compares to MX? It can't be as simple as divide by the surface area of the sensor, right?
 
It all looks amazin light wise and noise wise - also there is literally no skew in the longboard parkhouse scene that I can see.

but whats wrong with the car side shot @ frame 717 - 762? The upper line of the car shows some serious artifacts - not only the aliasing which looks really wrong but also color shifts just below the aliasing? whats going on here? What formula did you use to scale this - is the same visible in the original 4k footage?

Edit: Guess others had the same thought at the same time - but since they talk about the moiré and I talk about the aliasing I leave the question here - but both are probably caused by the same problem I guess.
 
It's most probably the 4:2:0 nature of the h264 codec used to distribute the movie, and that's half the reason why RED4k is a necessity. Detail like that is really pushing up against what traditional distribution codecs can handle.

Graeme
 
David Rasberry said:
If you look closely you will see a very fine crosshatch texture on the wheel surface around the logo. There will be moire near the Nyquist limit anytime two fixed patterns are overlaid, the pixel field and the surface texture in this case. You are seeing the Nyquist limit of 2k, not the camera original.
Click to expand...

To me this means the resampling algorithm did a horrible job. Would you say there should have been a low pass filter in there to fix that?

I think they should re-render the down-sample because A LOT of people are gonna get the wrong idea who are currently frustrated with the Canon DSLR Moire issues and not realize this is a workflow issue and NOT a sensor/debayer issue!
 
Graeme Nattress said:
It's most probably the 4:2:0 nature of the h264 codec used to distribute the movie, and that's half the reason why RED4k is a necessity. Detail like that is really pushing up against what traditional distribution codecs can handle.

Graeme
Click to expand...

Well I definitely cannot argue with that. I've spent some time converting footage from various formats into H.264 with a variety of options and have always been amazed at some of the issues it has.

I wouldn't be opposed myself to installing a different codec to see this footage with less artifacts myself.

And I'd like to SERIOUSLY add that the absence of large skewing or jello is absolutely awesome.
 
If I scale a prores to another prores then go to h264 using a 1:1 ratio with no resizing,
then the end result looks different to doing the whole thing in one step. Encoding h264 and resizing in one step
could have caused it, moire could have been caused by any number of things.

I highly doubt and sensor from red is delivering moire patterns in the raw
 
J Davis said:
If I scale a prores to another prores then go to h264 using a 1:1 ratio with no resizing,
then the end result looks different to doing the whole thing in one step. Encoding h264 and resizing in one step
could have caused it, moire could have been caused by any number of things.

I highly doubt and sensor from red is delivering moire patterns in the raw
Click to expand...

I certainly agree, I think someone just needs to fix that render! :) I only bring it up due to the point there are 3 characteristic sets people seems to fixate on with CMOS sensors (for good reason)

1. Moire / Aliasing
2. Sensitivity (DR, noise floor, etc)
3. Skew / Jello (Rolling shutter)

There are obvious other characteristics as well that are important, I am just mentioning what items seem to be the pain in most peoples sides. The Canon and Nikon DLSRs with video fail 1 and 3 pretty bad. They are very good for 2. Red, theoretically (and I say this because its not in the hands of the consumers just yet) excels at 1, 2 and 3. That said, I still want to see it :)
 
Got to say that if you're not watching the 4k in 4k for this reel, you're not seeing all the M-X is capable of. The new Sony projector we used to show it the other weekend was amazing, and the footage just shone!

Graeme
 
Rick Burnett said:
I certainly agree, I think someone just needs to fix that render! :) I only bring it up due to the point there are 3 characteristic sets people seems to fixate on with CMOS sensors (for good reason)

1. Moire / Aliasing
2. Sensitivity (DR, noise floor, etc)
3. Skew / Jello (Rolling shutter)

There are obvious other characteristics as well that are important, I am just mentioning what items seem to be the pain in most peoples sides. The Canon and Nikon DLSRs with video fail 1 and 3 pretty bad. They are very good for 2. Red, theoretically (and I say this because its not in the hands of the consumers just yet) excels at 1, 2 and 3. That said, I still want to see it :)
Click to expand...

Rick,

we could ask them to show us 4K JPEG of that particular frame (or TIFF, DPX) because there we can see is that an issue or not.
 
Graeme Nattress said:
Got to say that if you're not watching the 4k in 4k for this reel, you're not seeing all the M-X is capable of. The new Sony projector we used to show it the other weekend was amazing, and the footage just shone!

Graeme
Click to expand...

Maybe something to do at Red Studios. But I'd be happy with being able to DL in 1080p/2k. :ihih:
 
David Rasberry said:
If you look closely you will see a very fine crosshatch texture on the wheel surface around the logo. There will be moire near the Nyquist limit anytime two fixed high freqeuncy patterns are overlaid, the pixel field and the surface texture in this case. You are seeing the Nyquist limit of 2k, not the camera original.
Click to expand...

With proper downsizing this should not occur.
 
4k in all honors but I would say that a lot of people wanting a cam with this sensor are probably finalizing in 1080p/h264 for a very long time to come (blueray&web delivery) so it should look good there as well ;) just saying.

btw in apples shake you can choose which rescaling algorythm you want to use - maybe that helps in deciding which looks best.
 
Rob Ruffo said:
Yet people still buy F35s!? I don't get it.
Click to expand...

Because it's SONY. There is nothing better than a SONY camera. Everyone knows that. If it's digital or if it's video, you buy a SONY. Because nothing is better than a SONY and SONY is the best. All because SONY makes the most advanced digital video cameras in the world, at any price. If you want the Rolls Royce of digital video or cinema cameras you buy a SONY.

...Brains. Must, eat, BRAINS!!!!


Or perhaps I should have just said, "MMMmmmooooooooooooo..."
 
You must log in or register to reply here.
Back
Top