Chris Wright
New member
- Joined
- Mar 4, 2016
- Messages
- 3
- Reaction score
- 0
- Points
- 0
first off, I've read Adobe's whitepaper on native r3d imports and I believe it has errors....
Namely,
interpreting r3d as rec. 709 and a working space Rec. 709. THAT would lose quality from REDLogFilm during the keying process because you need the same size or bigger colorspace to work in. Instead I say, use view simulate to Rec. 709 and instead, process out as REDLogFilm gamma at color management DCI-P3/Redcolor4/bt1886(Rec 2020).
So for an ultimate keying guide, this is what I have so far... please annotate any working bugs!
step 0 shoot REDLogFilm it follows the cineon spec precisely, +-0.5 f-stop, 60 Ire, Green as 18% gray, 5000k not tungston. unlike other methods, proxies will not work with keying. Native ISO of the MX sensor is 800, old Mysterium-X RED One is ISO 320, ISO greater than > ISO3200 there's some mild compression to fit all the code values in = quantization. under Tungsten use a 80c blue Filter but loses 1 f-top
1. in redcine-x, do a FLUT white-balance point native D60, iso pull( so NLE doesn't need to pull shadows) while adjusting nothing else, denoise, Full debayer, REDLogFilm, Redcolor4.
2. in AE, chose DCI-P3/Redcolor 4/bt1886(Rec 2020) as the project working space and linear light for gamma 32bc, view-simulate rec. 709
3. import footage directly into AE as r3d, interpreted as REDlogFilm, Cineon converter log to lin
4. remove noise with neatvideo, sharpen off
5. put on keylight, increase green saturation AFTER keylight to view your work with use hue/saturation
if it looks good( just the alpha, not the color), then continue to next step.
6. Ignore spill suppression and color correction until later
7. use previous comp to alpha matte as a clean plate. Final step spill suppression, basic color correction, lightwarp etc.
8. render out REDlogFilm, colorspace DCI-P3/REDcolor4 or 64 cube lustre Lut redlog to bt1886(Rec 2020)
a. Tiff trillions color, cineon converter lin to log
b. 16 bit DPX Log
c. EXR linear light(log not needed in 32bpc space)
Namely,
interpreting r3d as rec. 709 and a working space Rec. 709. THAT would lose quality from REDLogFilm during the keying process because you need the same size or bigger colorspace to work in. Instead I say, use view simulate to Rec. 709 and instead, process out as REDLogFilm gamma at color management DCI-P3/Redcolor4/bt1886(Rec 2020).
So for an ultimate keying guide, this is what I have so far... please annotate any working bugs!
step 0 shoot REDLogFilm it follows the cineon spec precisely, +-0.5 f-stop, 60 Ire, Green as 18% gray, 5000k not tungston. unlike other methods, proxies will not work with keying. Native ISO of the MX sensor is 800, old Mysterium-X RED One is ISO 320, ISO greater than > ISO3200 there's some mild compression to fit all the code values in = quantization. under Tungsten use a 80c blue Filter but loses 1 f-top
1. in redcine-x, do a FLUT white-balance point native D60, iso pull( so NLE doesn't need to pull shadows) while adjusting nothing else, denoise, Full debayer, REDLogFilm, Redcolor4.
2. in AE, chose DCI-P3/Redcolor 4/bt1886(Rec 2020) as the project working space and linear light for gamma 32bc, view-simulate rec. 709
3. import footage directly into AE as r3d, interpreted as REDlogFilm, Cineon converter log to lin
4. remove noise with neatvideo, sharpen off
5. put on keylight, increase green saturation AFTER keylight to view your work with use hue/saturation
if it looks good( just the alpha, not the color), then continue to next step.
6. Ignore spill suppression and color correction until later
7. use previous comp to alpha matte as a clean plate. Final step spill suppression, basic color correction, lightwarp etc.
8. render out REDlogFilm, colorspace DCI-P3/REDcolor4 or 64 cube lustre Lut redlog to bt1886(Rec 2020)
a. Tiff trillions color, cineon converter lin to log
b. 16 bit DPX Log
c. EXR linear light(log not needed in 32bpc space)