Obins hacked 4:4:4 dvx clip
- Thread starter kyledroid
- Start date
seejay1031
Well-known member
Looks pretty damn good to my eyes. If I was looking to blowup to 35mm or High Def I would have to seriously consider this. Does anyone know what this could do in interlaced mode? The real reason I ask is the potential for full resolution 60 fps slow motion in SD.
Shaw- while not ideal lighting I bet you could pull a nice key off of the scene with the girl in front of the green wall. Kind of worse case scenario lighting situation with best case possible compression. Of course anyone using this mod would of course light GS much better.
Shaw- while not ideal lighting I bet you could pull a nice key off of the scene with the girl in front of the green wall. Kind of worse case scenario lighting situation with best case possible compression. Of course anyone using this mod would of course light GS much better.
J.R. Hudson
Spirit Animal
I second what is being said. It looks great. Impressive Latitude and Dynamic Range of Color (I'm not trying to sound like an asshole). A Split Screen Side by Side would be Epic.
That would be great John!
seejay - excellent question. I'd love to know about the interlaced footage as well... that's really the only thing that keeps me from jumping on this. I REALLY want true 1080p and 60p at 720 for slow motion. In the end I'll probably go for the HVX and perhaps have that moded somewhere down the line.
Good point about the greenscreen. Perhaps I should have said "uncompressed greenscreen"
. Still I bet I could get some decent results with this footage... hmmm. Off to test I suppose!
seejay - excellent question. I'd love to know about the interlaced footage as well... that's really the only thing that keeps me from jumping on this. I REALLY want true 1080p and 60p at 720 for slow motion. In the end I'll probably go for the HVX and perhaps have that moded somewhere down the line.
Good point about the greenscreen. Perhaps I should have said "uncompressed greenscreen"
. Still I bet I could get some decent results with this footage... hmmm. Off to test I suppose!
jbproflm
Member
Wow, that's the smoothest video image I've seen in a very long time. (Especially coming from a sub $7k camera setup.) The scene with the mouse and USB key is super clean. The dynamic range in these scenes is pretty intense. I will definently continue to keep my eye out for this mod.
thisiswells
Retired
This clip generally proves my point: Color correction in post is better than in camera.
They should modify the camera to record these kind of images to tape. 4:4:4 is nonsense. Just gimma the raw output of the camera, without any DSP, compressed in DV. Then, you are brilliant.
-Brian
They should modify the camera to record these kind of images to tape. 4:4:4 is nonsense. Just gimma the raw output of the camera, without any DSP, compressed in DV. Then, you are brilliant.
-Brian
Hate to disagree with Brian's post since he's usually consistantly right, but--what the heck--I'm going to disagree with it completely.
Shooting in RAW mode with my digi-rebel xt has spoiled me and I'll never shoot in JPEG again except for very casual shots. You're right that the data entering the DSP is has way more dynamic range than what leaves it but most of all what you're saying about color correction is true. I can totally manipulate white balance, exposure, etc. in ways that would leave artifacts in JPEGs and polarization in 8bit TIFFs.
But DV tape is not a capable format for "raw output." RAW mode on digital cameras is 12 bit uncompressed. DV is 4:1:1 "less than 8 bit" compressed at 5:1. Let's say you did capture all the 8 stops of dynamic range the ccd might theoretically have on dv tape by using a more advanced DSP. There would be less data per stop (8 stops going into 8 bits instead of 5-6 going into 8 bits) so color correction would result in much heavier posterization. The image would look beautiful coming out of the camera, but color correction would be a nightmare. It's like a DVD: sure the image has dynamic range really close to the 35mm film it started as, but if you try to color correct it, it's a mess. People who've used dynamic-range increasing ultracon filters have echoed this sentiment.
IMO, color correction IN camera is better than out of camera, assuming you don't overdo it. The DSP is 12 bit, the NTSC DV signal is less than 8 bit. So there's going to be way less posterization, etc. etc. in the DSP than on my computer. Working with these 10 bit TIFFS however (which don't have all the gain the DSP naturally applies, which is a huge pain), you have the extra dynamic range AND the extra data to manipulate it. The fact is that the DV format has certain limitations. Certainly you can get a good image from it, but the more you manipulate that image, the more its limitations show through.
Moral of the story: bring on 10 bit HD. THEN editing raw camera data will be a reality.
Shooting in RAW mode with my digi-rebel xt has spoiled me and I'll never shoot in JPEG again except for very casual shots. You're right that the data entering the DSP is has way more dynamic range than what leaves it but most of all what you're saying about color correction is true. I can totally manipulate white balance, exposure, etc. in ways that would leave artifacts in JPEGs and polarization in 8bit TIFFs.
But DV tape is not a capable format for "raw output." RAW mode on digital cameras is 12 bit uncompressed. DV is 4:1:1 "less than 8 bit" compressed at 5:1. Let's say you did capture all the 8 stops of dynamic range the ccd might theoretically have on dv tape by using a more advanced DSP. There would be less data per stop (8 stops going into 8 bits instead of 5-6 going into 8 bits) so color correction would result in much heavier posterization. The image would look beautiful coming out of the camera, but color correction would be a nightmare. It's like a DVD: sure the image has dynamic range really close to the 35mm film it started as, but if you try to color correct it, it's a mess. People who've used dynamic-range increasing ultracon filters have echoed this sentiment.
IMO, color correction IN camera is better than out of camera, assuming you don't overdo it. The DSP is 12 bit, the NTSC DV signal is less than 8 bit. So there's going to be way less posterization, etc. etc. in the DSP than on my computer. Working with these 10 bit TIFFS however (which don't have all the gain the DSP naturally applies, which is a huge pain), you have the extra dynamic range AND the extra data to manipulate it. The fact is that the DV format has certain limitations. Certainly you can get a good image from it, but the more you manipulate that image, the more its limitations show through.
Moral of the story: bring on 10 bit HD. THEN editing raw camera data will be a reality.
thisiswells
Retired
I'm glad you responded. To clarify, I have a fairly low opinion of CineGamma software. I would be interested in seeing the raw ccd output, compressed to dv, recorded on to tape. I believe I could manipulate even a highly compressed non-white balanced, non-color corrected, non-gamma curved image with better precision using tools on my Mac better than the simple software built into the camera. For what it's worth, Barry recently measured 7 - 7.5 stops dynamic range on his camera. It's in another thread. I don't really think it's the 4:4:4 data that is providing all the perceived dynamic range improvement. It's that they are using more capable tools in software to achieve those results. Before you say it's all because of uncompressed, full color resolution data that enables this. Try this little experiment: download one of the RAW files, recompress it to a DV frame (with appropriate JPEG file size) and try enhancing it in post. I think you will be surprised. Just a few random musings.
Well argued, but remember that there's also much less gain here because the DSP is pushing the footage at least a couple stops while these aren't pushed much at all. That said, if it's pushing the footage a few stops, it should be gaining even a few more stops in the highlights, as well, while it's instead losing them. I agree the DSP, while pretty good, could be a lot better. I also think that the hvx, if done right, could provide even better results than we've all been hoping for because the format is so much better.
thisiswells
Retired
To amend my prior post, a better way to test would be to export a raw frame as a DV clip, re-import, and color correct from the DV clip (as opposed to a Jpeg file).
I agree. The overall gain is the problem as you've pointed out. The DVX applies an appauling amount of gain before you even touch the "gain" switch. If it were possible to record an image to tape without any camera processing, I would actually pay 3 grand to have it. Of course, it isn't possible without building an entirely different camera... Or, as you say, the AG-HVX200.
I would imagine they have learned a lot in three years since the DVX100 and hopefully with offer a much more sophisticated algorithm in the new camera.. Jan Crittenden hinted on here earlier this year about the, "not clipping of the highlights" and this leads me to believe they have something up their sleeves in the form of better CineGamma that will continue underexposing by a couple of stops, but instead of cranking the gain across the entire image, it will be selective. Effective results would be a much improved dynamic range over the DVX100, regardless of the increased pixel density on the CCD's. This kind of innovation would make my day.
There is no "insider" information here, just keeping my hopes up, that's all.
I agree. The overall gain is the problem as you've pointed out. The DVX applies an appauling amount of gain before you even touch the "gain" switch. If it were possible to record an image to tape without any camera processing, I would actually pay 3 grand to have it. Of course, it isn't possible without building an entirely different camera... Or, as you say, the AG-HVX200.
I would imagine they have learned a lot in three years since the DVX100 and hopefully with offer a much more sophisticated algorithm in the new camera.. Jan Crittenden hinted on here earlier this year about the, "not clipping of the highlights" and this leads me to believe they have something up their sleeves in the form of better CineGamma that will continue underexposing by a couple of stops, but instead of cranking the gain across the entire image, it will be selective. Effective results would be a much improved dynamic range over the DVX100, regardless of the increased pixel density on the CCD's. This kind of innovation would make my day.
There is no "insider" information here, just keeping my hopes up, that's all.
Leigh Wanstead
Active member
Hello there,
I am lost here. Can anyone tell me how to get 4:4:4 from camera? I mean step by step, simple description.
I don't own the same camera, but I guess most minidv should get some sorts of similarity, so I can use the same method apply to my jvc gy-dv5000 and panasonic gs400.
TIA
Regards
Leigh
I am lost here. Can anyone tell me how to get 4:4:4 from camera? I mean step by step, simple description.
I don't own the same camera, but I guess most minidv should get some sorts of similarity, so I can use the same method apply to my jvc gy-dv5000 and panasonic gs400.
TIA
Regards
Leigh
Loki
Veteran
Thisiswells,
you say "Just gimma the raw output of the camera, without any DSP, compressed in DV. Then, you are brilliant."
correct me if I wrong here.. but i was under the impression that the DV compression was far more of a detriment to the image with the DV compression than the DSP...
but... If the DSP is what causes the loss in latitude and the NTSC conforming then yes I agree you are right and that is a good idea.. however .. if the Lattitude loss and NTSC conforming comes in with the DV Compression then going straight to DV wouldn't do to much..
the more I think about it though... what your saying makes sense..
does anyone have a flow chart of the image deterioration as it goes through the camera?
you say "Just gimma the raw output of the camera, without any DSP, compressed in DV. Then, you are brilliant."
correct me if I wrong here.. but i was under the impression that the DV compression was far more of a detriment to the image with the DV compression than the DSP...
but... If the DSP is what causes the loss in latitude and the NTSC conforming then yes I agree you are right and that is a good idea.. however .. if the Lattitude loss and NTSC conforming comes in with the DV Compression then going straight to DV wouldn't do to much..
the more I think about it though... what your saying makes sense..
does anyone have a flow chart of the image deterioration as it goes through the camera?
monte
Bag of Douche
To the best of my knowledge it works like this:
Image hits the lense - that's the first part that will degrade the image
Goes through CCD's
Then that image goes through the cameras internal things (Pedestal, Cinegamma, etc)
Finally, the image goes through compression and is written to mini-dv tape
So the way i count it that's four layers of image degradation. I suppose thats also the argument some people use if they say, "Getting the look in camera is better!" well I say, screw that, have it clean as possible then work in uncompressed in post for color correction. You might get some more artifacting and stuff, but it generally is fairly easy to fix and not too noticable. In my opinion.
Image hits the lense - that's the first part that will degrade the image
Goes through CCD's
Then that image goes through the cameras internal things (Pedestal, Cinegamma, etc)
Finally, the image goes through compression and is written to mini-dv tape
So the way i count it that's four layers of image degradation. I suppose thats also the argument some people use if they say, "Getting the look in camera is better!" well I say, screw that, have it clean as possible then work in uncompressed in post for color correction. You might get some more artifacting and stuff, but it generally is fairly easy to fix and not too noticable. In my opinion.
235 Studios
Veteran
Leigh Wanstead said:
Nope, that is a trade secret of reel-stream. Basicly what they have done, is take apart a DVX camera and introduced a way to capture the image straight off of the camera before it is compressed. In order to have this function on your camera you have to send your camera to them, and they mod it. Current;y only the DVX is being supported, and it is still in development. They have said that they plan on supporting other cameras in the future ... hope that helps.
Leigh Wanstead
Active member
waltind said:
Thanks
Regards
Leigh
