is there any way to import still images or videos in the VSE and export them (losslessly) without any processing done, so that the exported images are identical to the imported ones?
For instance, let’s say I want to import a dpx sequence and to export it as tiff or ffv1/mkv or any lossless format.
If I just use blender’s default settings, the exported images won’t be the same as the original, as can be verified checking the framemd5 with ffmpeg. This is not surprising as some default settings modify the picture : the color settings on “filmic” instead of “standard” for instance, dither… I know a few of these settings but not all of them.
So, which settings should I modify so I can have identical framemd5 upon export? In other words, how to get rid of any unwanted default image processing used in blender?
Dpx > non-dpx can not go through unmodified as a file, meaning file checksum will never start to match. Image data can, but then you have to do checksum on raw raster data itself, not files.
EDiT: just realized you did image checksum… set color management to blenders old default, not filmic (was it called standard or something similar, not behind desk…), it should generally bypass color changes. For mathematically lossless formats one thing that can modify data is bit depth, you must go to same or higher to avoid loss, but even then, if it goes through floats there could be type conversion related modifications from rounding.
Yes, I did set the color management on standard (that’s what I was I referring to by color settings in my post), also got the bit depth right. The framemd5 is still different though. I also tried setting the “dither” in post processing to 0, but the images are still not identical. There must be other settings to enable / disable but I have no idea which ones…
DPX support in Blender is a broken mess. Avoid it.
For display referred work, set the input file colour encoding to whatever it is encoded as. At that point, change the output view transform to the display linear equivalent of “Standard”.
Sorry, I’m not too sure I’m getting your solution… where are the menus in which the input file colour encoding (is that the colour management tab?) and the output view transform can be configured?
It is scattered around sadly. One place is Image Viewer in the sidebar.
Output rendering transform is per display, in the Colour Management Properties panel. So for example, if one is on an sRGB-like display, select Display “sRGB”, and the appropriate view transform. For the display linear no-operation, that would be “Standard” or “default”, which will encode the output with the inverse of the sRGB inverse EOTF, resulting in display linear range 0.0 to 1.0 being a no operation.
For radiometric-like footage, such as that generated by default in rendering engines such as Cycles, or properly encoded camera footage such as ARRI Alexa, Canon, Sony, Fuji, Nikon, Panasonic, in one of their [Vendor]-Log / [Vendor]-Gamut encodings, the generic display linear response will not work for properly interpreted footage. Using the default hardware display rendering transform is completely incorrect, and a more complex rendering transform is required. One such option would be Filmic, which is listed under Views.