DSLR Pink Color Shift-ProRes OBS Video
- Thread starter Mark Am
- Start date
Just a follow up in case anyone else runs across the same issue. It seems no matter what I set my source settings and advanced settings to, I am not able to eliminate the pink/magenta shift in the OBS recordings (compared to the H264 in-camera recordings):
Source 709/Limited, Advanced 709/Limited; Source 709/Full, Advanced 709/Full, Source 709/Full, Advanced 709/Limited, Source 709/Limited Advanced 709/Full [this not only resulted in a color shift but a noticeably more contrasty clip]; Source 709/Full, Advanced 709/Limited (nit raised to 3000); Source 709/Full, Advanced SRGB/Limited (3000 nit).
I don't know if its OBS itself or the ProRes software encoding that is causing the shift. The HDMI feed to OBS is 8-bit 4:2:2 and I'm using a good HDMI 2.0 cable. Anyway, in case anyone tries this at home, beware the slight magenta shift when recording to a prores file.
Source 709/Limited, Advanced 709/Limited; Source 709/Full, Advanced 709/Full, Source 709/Full, Advanced 709/Limited, Source 709/Limited Advanced 709/Full [this not only resulted in a color shift but a noticeably more contrasty clip]; Source 709/Full, Advanced 709/Limited (nit raised to 3000); Source 709/Full, Advanced SRGB/Limited (3000 nit).
I don't know if its OBS itself or the ProRes software encoding that is causing the shift. The HDMI feed to OBS is 8-bit 4:2:2 and I'm using a good HDMI 2.0 cable. Anyway, in case anyone tries this at home, beware the slight magenta shift when recording to a prores file.
I'm following up - again in case anyone else ever experiences this. As it turns out the difference between the internal Rec. 709 DSLR recording and the ProRes recordings from OBS was due to the ProRes recordings being slightly more saturated across the board - across all colors, not just magenta as I had originally thought. Once I dialed back saturation in Davinci Resolve the colors from the ProRes recordings and the DSLR internal recordings were nearly identical on the vector scope.
Long story: As it turns out I have a multicolored floor lamp in my office where I'm taking the test clips and coincidentally a few of the lamp colors nearly match the primary colors on the vectorscope in Davinci Resolve. When I compared the OBS ProRes clips with the internally recorded DSLR clips, the color patterns on the vectorscope reflecting the lamp colors were almost identical, except that the colors on the ProRes clips extended out a little further on the vectorscope than they did on the DSLR recorded clips. All it took was me dialing back the saturation setting on the primary wheels panel of the ProRes clips from 50.00 to 47.40 for all the colors reflecting the lamp to line up almost exactly where they were on the DSLR-internally recorded clips.
I'm wondering if there is a record setting in OBS that is causing my ProRes Rec 709 clips to be slightly more saturated than the 709 clips recorded directly from the DSLR, or if that is just to be expected with ProRes recordings.
Long story: As it turns out I have a multicolored floor lamp in my office where I'm taking the test clips and coincidentally a few of the lamp colors nearly match the primary colors on the vectorscope in Davinci Resolve. When I compared the OBS ProRes clips with the internally recorded DSLR clips, the color patterns on the vectorscope reflecting the lamp colors were almost identical, except that the colors on the ProRes clips extended out a little further on the vectorscope than they did on the DSLR recorded clips. All it took was me dialing back the saturation setting on the primary wheels panel of the ProRes clips from 50.00 to 47.40 for all the colors reflecting the lamp to line up almost exactly where they were on the DSLR-internally recorded clips.
I'm wondering if there is a record setting in OBS that is causing my ProRes Rec 709 clips to be slightly more saturated than the 709 clips recorded directly from the DSLR, or if that is just to be expected with ProRes recordings.
I posted my results in another thread, but figured I's post them here as well:
I did a few more comparisons between the H.264/709 file in camera and both ProRes and H264 files recorded in OBS. I shot the identical scene with that nice multicolored lamp in the frame each time. Long story short the OBS recordings were all nearly identical on the vectorsope (the tiny vectorscope differences were not detectable when viewing the clips on screen at least to my eye). So the 8-bit OBS H264/709 clip (using limited/limited settings if I recall) was nearly identical to the limited/limited and full/full 10-bit ProRes OBS clips. All three clips were just slightly more saturated than the internally recorded H.264/709 clip. Since the 8-bit and 10-bit OBS clips were nearly identical and they all required the same slight adjustment to match the internally recorded clip, I'm wondering if the difference in saturation is due to the 4:2:0 recording that the camera does internally vs the 4:2:2 stream that it puts out on the HDMI.
Anyway, this concludes my testing - I'm really happy to be able to shoot 10-bit ProRes files with that DSLR, understanding that if I want colors to match identically what would be internally recorded I just have to adjust saturation slightly.
I did a few more comparisons between the H.264/709 file in camera and both ProRes and H264 files recorded in OBS. I shot the identical scene with that nice multicolored lamp in the frame each time. Long story short the OBS recordings were all nearly identical on the vectorsope (the tiny vectorscope differences were not detectable when viewing the clips on screen at least to my eye). So the 8-bit OBS H264/709 clip (using limited/limited settings if I recall) was nearly identical to the limited/limited and full/full 10-bit ProRes OBS clips. All three clips were just slightly more saturated than the internally recorded H.264/709 clip. Since the 8-bit and 10-bit OBS clips were nearly identical and they all required the same slight adjustment to match the internally recorded clip, I'm wondering if the difference in saturation is due to the 4:2:0 recording that the camera does internally vs the 4:2:2 stream that it puts out on the HDMI.
Anyway, this concludes my testing - I'm really happy to be able to shoot 10-bit ProRes files with that DSLR, understanding that if I want colors to match identically what would be internally recorded I just have to adjust saturation slightly.
Suslik V
Active Member
Everything can be different. From the transfer function to media player mistakes (or it's just your imagination that cannot accept chroma subsampling).
If possible, try to upload test images to the camera and then capture it (yeah, some cameras can show images that were saved only with the software from the manufacturer of the device, but anyway give it a try).
For 8-bit videos some test charts (images) can be found in the https://obsproject.com/forum/resources/categories/guides-studio.8/ section of the forum. Make few custom files with different chroma subsampling (FFmpeg can do the job), and you'll understand how same image looks when its chroma was subsampled. Don't forget to switch your display device and OS back to 8-bits before performing tests...
If possible, try to upload test images to the camera and then capture it (yeah, some cameras can show images that were saved only with the software from the manufacturer of the device, but anyway give it a try).
For 8-bit videos some test charts (images) can be found in the https://obsproject.com/forum/resources/categories/guides-studio.8/ section of the forum. Make few custom files with different chroma subsampling (FFmpeg can do the job), and you'll understand how same image looks when its chroma was subsampled. Don't forget to switch your display device and OS back to 8-bits before performing tests...