jack jin

On 12/19/2020 at 10:25 PM, Sage said:

Lol @ the ending

Hi @Sage, I hope you've had a great new year so far, I have used the s1alex lut on various personal projects now, gotta say I love the colors that it gives me, much more consistent and pleasing results even in warm or weird lighting compared to the p4k version. But the exposure seems to be quite a bit brighter than the normal rec709 lut. And the lut shows a lot more noise and artifacts then I expected especially at the higher iso gain. I have tried to reduce the exposure using the exposure adjustment luts, but on darker shots it clips the shadows and doesn't roll them off naturally like the braw exposure adjustments. Any advice on how to reduce the noise and artifacts without using noise reduction? (PS: I am using tetrahedral 3d lut interpretation in Davinci Resolve already.)

Filmconvert nitrate already did it, with a match the cineon log scan of a bunch of filmstocks, of course there are still things missing like gate weave, accurate halation etc. But in terms of color, resolution, and dynamic range, something like the c70 already exceeds film.

21 minutes ago, Eric Calabros said:

Impossible is strong word for a technical limitation like that. 

I think it's just that a mirrorless mount ilke e mount simply do not have enough space for both the flip down electronic nd mechanism and an ibis mechanism.

@IronFilm  @Emanuel

2 hours ago, IronFilm said:

WHAT??? Why on earth would RED leave that out.  

Neither can IBIS be added with a firmware update.

Because red is trying to protect their higher end cinema line with a less flexible codec option, it's still named "redcode", but it's dct based like h.264 and prores. 

Sony's internal Electronic ND is unfortunately not compatible with ibis.

Computational photography is the future, no matter the sensor size. But one thing smartphones are still struggling to do is to deliver a natural looking image. Skintone look plasticy under certain lighting situations, and the aggressive denoising and sharpening isn't helping thing either. The alexa managed to do dual iso 10 years ago, while delivering the most organic digital image ever created. So when smartphone's are able to deliver that level of natural looking photos and video. I think then it's time for computational photography to be widely adopted across all platforms.

5 hours ago, Thpriest said:

Do you notice any difference in low light? Noise etc?

 

It does look great in low iso and with sharp glass, but it's not as sharp as the fullframe 6k downsampled to 4k where the detail just pops off the screen, in lowlight the noise is approximately 2/3 of stop noisier in my experience.

11 hours ago, Thpriest said:

Uffff! Great news! I was about to sell my 18 -135 tomorrow but I'll cancel the sale now! This would mean I could make the change to the L - mount without having to initially buy a lot of expensive lenses as I'm using it for 90% video and manually focus most of the time! If the peaking is up to scratch then I'm good to go.

Btw the autofocus is surprisingly a bit faster with the 18-135 compared to my tamron 24-70 g2, even though on a native ef camera the 18-135 is a lot slower then the tamron. Also apsc is limited 12mp, so the resolution you get out of your photos won't be great, same for 4k video, where it is only a one to one readout, so the noise and resolution isn't gonna be the best.

4 hours ago, Sage said:

@Emanuel I had heard that it was to use the GH5S sensor. If so, there is a reasonable chance that the color will be the same. If not, then it could be measured. I didn't know about the Osmo, but it is something that could be explored. There were some cool experiments I wanted to mess around with, so a deep dive may be warranted there next.

@HockeyFan12 Very cool, I haven't read that one yet. I did notice S1 moire on saturated fabrics, and in that moment felt the value of the S1H. Here are the S1 fabrics compared to the Alexa:

S1

Alexa

@SageWould you mind test grading 5 dpx in different lighting situations with the different color temperature luts for me? I found some footage online and on facebook. And would just like to see how the tungsten and 4200k variant holds up in difficult situations before buying the camera and lut package.

10 hours ago, mirekti said:

Just bought the package for S series.

It kind of pulls towards green cast a bit, and if I change tint towards magenta the lips become a tad pink. Any alternative method to try to fix that?

That's what I'm worried about as well, was this shot in daylight? and was it using the daylight lut?

34 minutes ago, HockeyFan12 said:

Interesting. That particular workflow is a bit over my head, but the cineon conversion reminds me of the cinny LUT here:

https://www.thebrim.pictures/sucomo.html

Seems really useful.

The linny lut is not strictly accurate, because you don't know what display prep they based the original emulation on.  It's a emulation of a scan of a print of a negative. There is so many variables that could change the colors drastically within their pipeline. What if the color timing was off during the printing process? Did they print the negative optically or through laser? Did the negative or print went through Digital Intermediate before the print or scan? What was the machine that scanned the print? Where as the filmconvert nitrate is just simply a conversion to the filmstock's cineon colors. Not to mention the linny lut is ridiclously expensive, with only one stock, where as filmconvert nitrate got 19 stocks, with options for grain emulation for 1/5th the price. And the cinny lut only includes one stock, which filmconvert has, while costing 395 bucks for one lut. It's absolute insanity how much they think they can charge for their luts.

1 hour ago, HockeyFan12 said:

*where do you go from the cineon conversion

Three nodes, third node you input the status m density rgb matrix shown here:

into davinci resolve, set the saturation to 60. On the second node you first put in filmconvert nitrate, select camera profile (must be log for cineon to show up), and then put print film emulation to 0, then you go into the white balance tab to adjust the white balance until it looks neutral, these matrix mess with the white balance so the image might look very magenta and warm, this is where you fix it. I like to use the classic rgb matrix, it gives me the best skintones. Going back to the third node, I just simply manually adjust the gamma/curve to my liking, by setting the black level and then applying a slight s curve to the footage, or I use CST to conform the gamma from cineon log to rec709 with luminance mapping turned on.

4 hours ago, HockeyFan12 said:

Will the S1 power grades work for EVA1 and Varicam?

Also is it possible to make these transforms into LUTs? Can't you export a power grade as .cube?

Would it be possible to do a more accurate "film look" this way? I don't understand why FilmConvert doesn't really look that good to me, but the Brim does a pretty good job to my eyes and Arri originally had a special Log C setting meant to emulate film scans. It was problematic and they discontinued it, but the colors were a perfect math to 5219 I think. Log C now has its own look, a bit more like reversal than color negative imo.

Have you tried filmconvert nitrate? The cineon log conversion has spot on colors and contrast, it's their normal emulation laced with their heavy print film colors that looks crap. 

1 hour ago, Sage said:

@jack jin Is it the Daylight conversion? This is common when Daylight is applied to tungsten lighting

@TomTheDP Sure, that's possible; I am at a crossroads, wherein there are some experimental ideas I've wanted to mess around with for years. Now with the engine complete, I feel more free to explore them

Ohhhhhh, so that's why? If I use tungsten version of the lut does the yellow hues and brown hues stays consistent in warmer lights and underexposed situations? Thank you for answering my questions!

11 hours ago, Sage said:

Apologies, I was away for a bit. An update - first, my dad is out of the hospital and recovering from surgery.  We are ostensibly through the worst of it.

Second, I ended up rewriting the complete engine. All legacy code has been replaced, in order to make extreme saturation projection more accurate, as well as make some improvements along the way (for example, the engine now features a 1D function at core)

Third, I've completed the S1/S1H/S5 conversion, as of yesterday.

It is the most recent LogC (SUP 11.1.1)

Fortunately the S series seems to have a better internal codec. The GH5 was uniquely problematic with its red/green macroblock splitting. To test gradient motion, film a white wall in darkness. Then, move a spotlight back and forth to see If there is banding revealed as the spotlight falls into darkness. Better codecs will handle this smoothly, while more compressed codecs may have some visible bands moving in the gradient. This is the kind of thing that can crop up in lens flares, or panning across gradients.

Thank you very much for the explanation! Happy that your dad is recovering! But I have noticed that in underexposed situation and in warmer lights the yellow hue with p4kalex shifts to green. Thus I wasn't able to use the lut in those kinds of situations. Is it simply caused the the p4k colorscience or is this common with all v4 color? Since when using the v3 ghalex all the brown and yellow hues have very nice colors to them. Thank you very much! 

Have anybody used the tokina 24-70 f2.8 with the mc 21? I would like to pair them up since that's one of the cheapest brandnew 24-70 in my country, and we don't have access to the fotodiox adapters, so if anybody have any experience with them it would be great if you can share about it. Thank you.

On 7/25/2020 at 9:13 AM, Sage said:

It should be excellent. The mode that I found problematic was actually the 4K 150mbps, specifically on a gradient motion test. 4K 400 and 1080p All-I were both very good here (as well as 1080p 100 long). I'm a 2K fan generally (typical minimalism), hence the preference for 1080p All-I (the GH5 has excellent down-sampling)

Oh no! What will this mean for the s5 and s1 which only has the 150mbps codec? Are they breaking apart very easily compared to the 400mbps all I? And what does gradient motion mean exactly? Is it macroblocking or banding? I am really looking towards to your s1/s1h/s5 version but I am worried about footage breaking apart since the s5 only has the 150mbps codec.

15 hours ago, Lux Shots said:

The S1/S5/S1H 4K 24p 10-bit is H.264, while the 4K 60p 10-bit is H.265. The S1H 5.9K 10-bit internal is H.265.

I haven't forgot about you, I have just been crazy busy. I'm also done with the Facebook group, too much bullshit there for me to be honest. I'm trying to set something up with a model, but models can be nervous when you ask them to shoot in low light at night for the first gig. I'm going to try and get some time on Wednesday for a photoshoot, and then maybe I can get a video shoot with her later in the week.

Haha holy crap dude, can't believe that you are still gonna do this. Thank you very much for the sacrifice bro. I also do wonder what the 4000, 6400 and 12800 looks like in vlog, so if you can maybe shoot some footage at those iso levels as well that would be amazing. Thank you so much!

2 hours ago, HockeyFan12 said:

Are you using Adobe or transcoding to ProRes with Apple software first? Adobe's interpretation of HEVC is broken. I can post comparisons if you're curious. I transcode to ProRes (in Finder, actually) first and it dramatically improves the HEVC image.

The footage that I'm finding this issue on are all transcoded and edited prores 422/hq sequence, because they are the only ones with underexposed and lowlight shots with a lot of skintones. And I'm pretty sure that the s1/s5's 10bit codec is h.264, not hevc.

4 hours ago, hyalinejim said:

Is it a  straight out of camera file?Yesterday I downloaded some "sample" S1 VLog  that was edited and encoded to ProRes. When I pushed it I saw highlight banding that I never see with GH5 VLog mp4s, and that I haven't seen with S1 SOOC files.

YEAH!!! It seems like all of the straight out of cameras were fine with this, but all of the prores transcoded stuff i downloaded (which contains way more underexposed and lowlight stuff) has slight to medium levels.of artifacting like this. But I'm not sure since I can't find any SOOC files in a underexposed situation or lowlight, if you or anyone has some vlog 10bit sooc files in those difficult situations with skintones that you can share that would be amazing.

Just wondering if any of you have seen some artifacts from your 10bit s1 vlog footage. From what I downloaded back then I am seeing color blotches on slightly underexposed skintones and surfaces, which makes me kind of worried about the s5's vlog footage since they are using the same codecs. Have anyone else experienced a similar artifact?

3 minutes ago, deezid said:

Not the reason at all.
Both Nikon and Sony do heavy processing on their footage. Strong NR creating all kinds of artifacts and strong sharpening with halos around corners. 

If you want a proper Sony camera which doesn't look like video, the FX9 and above are your only choices. While Panasonic starting with the S5/S1 and above looks really organic.

Interesting... Because from what I see in the dpreview comparisons: https://www.dpreview.com/reviews/image-comparison/fullscreen?attr29_0=panasonic_dcs1&attr29_1=nikon_z6&attr72_0=4k&attr72_1=4k&normalization=full&widget=200&x=-0.1766201071300748&y=-0.717123739236587

It seems like the s1 and the s1h is about as sharp as the panasonic gh4 which does a 1:1 pixel readout, even though lots of people claim that the 4k output out of those camera are downsampled from 5.9k. Where as the a73 and z6 looks quite a bit more detailed. And with the apsc xt3 blowing all of those cameras out of the water. But the image out of the s1 is indeed quite organic. The 10bit output does look great.

15 minutes ago, deezid said:

Just some slight spatial filtering on the S1H (NR-1), S1 and S5. Nothing which creates visible artifacts. 🙂

Is that the reason why the s1's video is so soft when compared to the a73 and z6? Does the spatial filtering filters out high frequency details as well?

7 minutes ago, deezid said:

-1 on the S1H equals 0 on the S5 and S1 😉
No issues with ghosting or smearing.

This is the vlog menu on the s1, and how about sharpening? Usually you can turn them both down to -4 so i don't see how this shouldn't be adjustable on the s5.

Does the s5 has the ability to adjust the Noise Reduction and Sharpening in VLOG? From what I seen in some videos it seems like those options are greyed out. Which would be a massive bummer since I hate the aggressive noise reduction and sharpneing from these mirrorless cameras.