Using a program like "Alien Skin blow up" on video layers in PS
Hi, I do some 3d rendering and generally output to a png image sequence. Because of frustrating long render times I often render in 720 p instead of what I would really like to which is 1080p
What I was wondering is if I could render an animation in 720p in my 3d program. Import the image sequence to a video layer, and then use a program within photoshop.. like eye candy Blowup, to upsize to 1080p.
Does anybody know if this would be possible?
Wow... 4 days and not even a guess from anybody? Maybe not many people have used PS's video features. I don't want to invest in the plug in, if I don't know for sure if it would work.
I sometimes use Photoshop for video and I have used a few of the "blow up" utilities in the past. Generally, they are meant to enlarge the output (to the printer) and not necessarily to enlarge an output file. You may be able to get them to actually output a larger image, but I'm not sure. Most of them have some sort of demo to download and try.
Bear in mind that the algorithms they use to ensure quality will require significant computer time, especially if you're processing a lengthy image sequence. You might want to get a feel for the quality and time involved by having PS up-scale an image sequence.
Another solution would be to use one of the "enlargers" for Premiere or After Effects. InstantHD comes to mid, but there are probably others.
However, I think you might be going about this in the wrong way. Uprezzing an image will certainly cause some quality degradation. Plus ANY sort of up-rezzing is going to be a bit of a compromise, especially regarding time versus quality. You're likely to get much better quality outputting a larger sequence from your 3d program. And I doubt that you will save significant time with a workflow that involves outputting 720p, importing into PS, uprezzing, then saving the file.
If you haven't found another solution already, be sure to try Detail Preserving Upscale available in Photoshop CC and in the very latest release of After Effects CC.
Pretty impressive results and no added cost.
For those who don't know:
When you upscale a single image for processing, it's USUALLY for a printer output, but OCCASIONALLY for a screen output. Eh.... ....that was about 10 years ago... ...Let's try again, shall we?
When you're customizing a single image to an output media, you have to remember a few things. First, PANDA. Pixels Are Not Dots. Printers use dots per inch as a measure. If you set your output resolution to match your pixels to your dots, remember this: If a dot is placed for every pixel, then the printer has to create the color for EVERY Pixel by placing ink on it, which can cause problems with some, where the color smears as the page moves. This has been a fight for so long, people still don't realize that printers actually account for this smudging to create the pixels' actual color variation and transition. So don't try to resample an image to fit your dots per inch, but rather about 1\2-1\4 of that, and you'll see significant improvement. I've had great results with values like 266ppi for a cheap 800dpi printer. PPI used to mean points per inch, to define the images used for text characters. With modern imaging, it has been able to stand for both points and pixels, depending on your application. If you use points, it's usually a square section of pixels on Long Pixels screen sizes (aspect ratios of 4:3 with rectangular pixels), or equivalent to 1 actual pixel (or a value like 4 or 9; a perfect square) on modern screens, based on your program's design. Modern programs can use 1pt per PXL, making the term interchangeable. This allows you to resample an image to a larger or smaller size by following simple steps to provide the transitional data, then stretching it. First uprez the PPI, let the program cut the pixels for you. Now blow up the image by a factor, and let the program resample at current ppi. Now set the PPI to your output media data size. Screens are usually 72-120ppi, video is 72. People say "You can't CREATE detail!" and shout it all the time. They are right from the perspective that you cannot rebuild what wasn't there to begin with. However, you can create a more pleasing transition, or a finer edge to add clarity and depth.
If you want to blow up your video by going frame by frame and blowing it up, DO NOT START WITH VIDEO DATA. Render the JPEG or TIFF version of each frame, work on one or two and find the common setting of best fit for the process described above. Your ultimate sizing should be 150ppi to 300. Many ask why when the video scales to 72ppi? Well... ...It's complicated. When resampling the images for video files, you want to start with as much info as possible, and drop down during the processing. This resampling will drop more data, but will keep more of your sharper edges and transition areas if there's more pixel data to use in the calculation. Once you have the settings, you can write them down, then create an action that plays. Apply this action to a whole folder of your video images. Now go and have fun for 6-24 hours, because this will take some serious time and processing power. If you know scripting and can do this in a script, sending yourself an email when it finishes the lot, great, do it and use that knowledge, baby! Now you can set it and forget it until your pants vibrate awkwardly on that special date.
Once you resize the images and save them to another location, they should be 2-4 times the data size. You'll need enough space for 5x the size of your UNCOMPRESSED (jpeg images) video. Keep the originals until you're done, this helps as a backup, but also provides some other possibilities.
Once you have your video frames, you can wrap them in a video or place them in a timeline for cutting (extremely slow playback); or you compress a working video from them and replace it with the non-compressed version later, which yields time saved and impressive detail retention.
Here are some options for you:
If you only blow up by one full step (i.e. 720x480 to 1280x720), you can employ a technique called Detail Matching. This technique requires the following--
1. You place your new large files as the odd frames, and your originals (simply scaled up) as even into a video file that is double your original frame rate
2. You output a file at your original frame rate, blending the two sets of frames together.
If you sharpened and applied some grain before upsizing, and retained a lot of sharp detail, this will smooth some out, making it more visually appealing, as the color and transition areas will blend and smooth slightly. It's flattering and visually appealing.
Another technique requires that you completely over sharpen and high-pass each frame, then create a DETAIL MAP from that (you create a black and white edge map, a bit like a sketch with the edges all very dark black and everything else perfect white), then you create a VECTOR DRAWING of said map, save it, and apply it to the blown up version of the original as a detail enhancement. Play with a few images and blending modes to get this close, then create an action and apply it (now you need 6-7x the space of your uncompressed video). By blending the vectorized details, you'll retain more of the detail of the image, scaled almost perfectly. Bring them back with subtlety, as it'll look awful with too much. Use this method and simply make an output video to work with in sequence. Again keep the ppi to 300.
Why a vector drawing? It resizes perfectly, even if it is a huge file. But with only 2 colors, one of them a 0000 value, it will be a little smaller, generally, saving a little space. Blowing it up will create natural breaks in some lines with rasters, but vectors calculate exact pixels and will connect them, keeping shapes and curvature a little more true. By blending the adjustment into your original, blown up, you'll be affecting areas that were smoothed over, and applying a sort of grainy sharpness back to them. Thats why you should do this with subtlety, as it can pixelate the image if not done carefully.
Of course, you can apply both methods. IF you get a grainy look from the second, do the odds\evens method using your originals, and the grain will smooth over. This doesn't CREATE detail, it simply duplicates existing micro contrast as a vector map, then applies the vector map to a larger frame. Because this is so easy to overdo, applying a resample-scale-smoothed version of the original as a blend provides a kind of smooth filler, pulling the effect back a bit.
This isn't an overnight kind of operation. It's DAYS upon DAYS of rendering and action building. It WILL NOT CREATE DETAILED UPRES of ANY video. IT WILL RETAIN MUCH MORE DETAIL WHEN UPRES-ING MOST video.
This is for those with varying video rendering setups, especially with older software packages with perpetual license ownership (you know, like CS6 or 5).
I can confirm some decent results with the detail retention of AFTER EFFECTS CC at it's current version. Premiere gets OK results, but AE gets excellent results. It does so by changing the PPI without changing the actual number of pixels in the image. IF I have a 6ppi image that contains 300 pixels total, it's inches are 50 effective inches. If I drop the sampling of the pixels to 3, I have 100 effective inches, but the detail in those 3 pixels per inch is drastically smoothed over. This is where resampling takes over and recalculates color values in the pixels. When done this way, some colors are blended to create an easier transition area, with smooth edges. If you then alter the pixel count of the image, keeping the rest of the sizing normal (inches), you can raise the PPI back up, dropping the effective size in inches to get back to the original output PPI value, with a larger image size. This stretches and smoothes the transition areas a little more, and squeezes some areas together creating hard edges. Afterward, a sharpening algorithm applies more contrast to edge areas (unsharp masking with a high threshold, and narrow radius). Where there's already a hard edge, the threshold applies some drop in contrast\noise to reach the average value, and viola. You have a clean edge, sharpened for screen use, and blown up to a larger output size. That's essentially the way the plug functions. However, in premiere, the engine is less precise with values and it will only work about one step up. After effects can pull 720p to full 1080 or 3k with relative clarity. I'm told it incorporates a High-pass filtering that maps details and blends them back in.
thanks, @ht davis. you actually know your stuff! i've spent 3 hours this morning to figure out the best upscaling methods - as common as the task is, the web is full of ignorance and outright misinformation.
yes, ae detail preserving does OK for me - as did instant HD advanced in the past (discontinued, even if its algorithm had something others don't - pick add'l resolution from the neighboring frames).
for image sequence, maybe alien skin blow up will do - and it automates some of the steps so you don't end up in computing limbo. trying it now.