A possible workflow for the GoPro Hero4

GeeBax wrote on 11/2/2014, 3:05 AM
Some of you may recall some time ago I bought a Sony AS100VR for use in my work, primarily because I wanted the image stabilisation feature. Unfortunately I was not happy with the camera due to the severe fisheye distortion and the odd clumpy video artefacts that made the pictures poor in apparent resolution.

On the fortunate side, I was able to put the AS100 up on EvilBay and got a top bid that was only 6 bucks under what I paid for the camera (don’t you just love it when people get carried away with bidding), so I then ordered one of the new GoPro Hero4 Black cameras. There was not a lot of material around to judge the Hero4 by, as the camera was quite new, and this made me a bit nervous. But the small amount I saw led me to believe it would be reasonably good for the price, even though it was a couple of hundred more that the AS100.

I now have the camera and have been performing tests with it to try and develop an optimum workflow for processing it. My target is to produce 1080p HD material that is crisp and sharp, so the Hero 4 material will be mixed with footage from my Blackmagic Pocket Cinema Camera, and I would like to get them to look as close as possible when cut together.

With that in mind, I shot a test clip in 2.7K 24fps using the GoPro Protune settings to obtain a clip that was as unprocessed as possible. To obtain the footage, I attached the camera to the front of my car using a suction mount on the hood and drove from my house down to the nearby town of Garfield and back. This is a close approximation to the sort of shooting I want the camera for, where it will be mounted on moving railway vehicles, often quite low down.

The Hero 4 offers 4 setting for the FOV, SuperView, Ultra Wide, Medium and Narrow. The 2.7K resolution setting offers 3 of these, omitting the narrow FOV. I chose the medium setting as I wanted to begin with the least distorted image possible. I then tried processing the footage through the GoPro Studio application, a free download authored by Cineform. GoPro Studio has a function to remove the fisheye distortion, and even though it was not particularly objectionable, I thought it might be worth a try. The result was not very good, it cropped the image quite a bit, reduced the resolution of the image and more importantly, totally cancelled all the Protune settings. I rather suspect the GoPro Studio application is not for the discerning professional but most likely aimed at the GoPro fans who have little idea about good image processing.

So I turned to DaVinci Resolve 11 Lite, which is a free download. I was able to import the footage easily into Resolve, and then graded the footage lightly (the color was close to ideal as it came out of the camera). I restored blacks, adjusted the gamma, increased overall saturation a touch and just added a small degree of sharpening to the image. I then rendered the footage as a uncompressed 10 bit YUV file, and noted the growth in file size. The original camera footage was about 1.7GB, it grew to 42GB when rendered. Under normal circumstances I would then use this file to edit in Vegas and perform all titling and audio work. However as this is a test only, I then processed the file through trusty Handbrake to arrive at a file size around 578Mb. The resultant file is a compromise between keeping the file down in size and the appearance of ‘clumping’ in the tree foliage. For my final work, the size will not be a problem.

My conclusions are these. I will probably not attempt to correct the fisheye distortion, as I do not find it that objectionable if I shoot in ‘Medium’ FOV. I may increase my rate to shoot at 30fps to try and minimise movement artefacts. And I will use Resolve to do the grading and output scaling (its free and works a treat). I think the Hero4 is a very good investment, even though the cost is getting a little high, but the pictures prove it is worth the extra money. My only beef is the fact that it blows out the highlights a little too much for my liking, but all the other automatic features make it a winner for this sort of operation where you are not able to control all the parameters during shooting. The WiFi capability of the Hero4 was a pain to use, consumed lots of battery and frankly I will not bother with it again. I tried the Wifi using the GoPro App on my iPad and came close to throwing the iPad over the fence. I really only wanted to be able to see the framing when setting it up, but in the end I cut a hole in the side of the GoPro waterproof housing and plugged in my 7” viewfinder via the micro HDMI connection, much better result.

I was investigating purchasing the Backbone Ribcage conversion that allows you to remove the GoPro lens and fit a mount for C and CS lenses, but the combined cost of this conversion plus a suitable lens amounts to around the same price as another Hero4, and having sampled the footage using the ‘Medium’ FOV setting, I am now inclined to drop that idea.

Would I recommend the Hero4? Certainly, it is a very worthwhile tool to have in your armoury. Best way to process the Hero4 files? Resolve Lite, absolutely.

If you would like to see the clip after processing, you can download it from here, it is Clip 5 at the bottom of the list: http://www.hollywoodfoundry.com/film%20files.htm

But be warned, it is 578Mb in size. I do welcome critical discussion on the clip and my findings.

Geoff

PS, my next experiment will be to explore the stabilisation feature in Resolve 11, although this clip hardly needs it.

Comments

NickHope wrote on 11/2/2014, 3:12 AM
Geoff, how steep was the learning curve in Resolve Lite?

Did you watch or read any tutorials, and if so, can you recommend any?

Is there no way to render out to lossless or near-lossless compressed formats in Resolve Lite?
NormanPCN wrote on 11/2/2014, 9:53 AM
The result was not very good, it cropped the image quite a bit, reduced the resolution of the image and more importantly, totally cancelled all the Protune settings. I rather suspect the GoPro Studio application is not for the discerning professional but most likely aimed at the GoPro fans who have little idea about good image processing.


What crop did you get? Was the output set to 1080? With my Hero3 I have used 2.7k and gotten a 2.7k Cineform output from the free GoPro studio.

What do you mean by cancelled all the Protune settings? Protune are just settings the camera uses to encode the video file. The file is what it is and set in stone.

The free GoPro studio does automatically apply a "Protune" preset to a Protune source file that makes the file look like a "normal" file with regards to contrast and color. You do not have to keep that. You can click the None preset and then tweak as you wish.

[edit]All that said, I normally just bring my GoPro files directly into Vegas.
johnmeyer wrote on 11/2/2014, 11:14 AM
Have you tried to duplicate any of the shots you took with your AS100V? You may remember that I posted similar negatives comments to those you just recounted. In your review above, while it is clear that you are happier with the fisheye, it wasn't as clear whether you found the detail and compression artifacts in the Hero 4 significantly better than the AS100V.

So, is the picture quality better, and can you provide any examples, possibly including some AS100V footage for comparison. Even still photos would be useful.
GeeBax wrote on 11/2/2014, 3:36 PM
[I]Geoff, how steep was the learning curve in Resolve Lite?[/I]

Nick, I had done many years of colour grading in my professional life, so I did not find it all that difficult, but on the other hand, some of the functions in Resolve are not particularly intuitive. There is a very good on-line manual for it though and from then on it is grope-and-hope :-)

Resolve offers a large range of output formats, I generally use Quicktime Uncompressed YUV 4:2:2 10 bit.

[I]What crop did you get? Was the output set to 1080? With my Hero3 I have used 2.7k and gotten a 2.7k Cineform output from the free GoPro studio.

What do you mean by cancelled all the Protune settings? Protune are just settings the camera uses to encode the video file. The file is what it is and set in stone.

The free GoPro studio does automatically apply a "Protune" preset to a Protune source file that makes the file look like a "normal" file with regards to contrast and color.[/I]

I did not actually output the file from GoPro Studio, simply used the import process with the 'Remove Fisheye' option selected. It cropped the image to a degree, the image became softer and as you mention, it appeared to apply a preset to the file, none of which I liked. I could not find any option to disable the preset in the import process. I did not bother to output the file, as the imported (intermediate) file was not to my liking.

[I]Have you tried to duplicate any of the shots you took with your AS100V? You may remember that I posted similar negatives comments to those you just recounted. In your review above, while it is clear that you are happier with the fisheye, it wasn't as clear whether you found the detail and compression artifacts in the Hero 4 significantly better than the AS100V.

So, is the picture quality better, and can you provide any examples, possibly including some AS100V footage for comparison. Even still photos would be useful.[/I]

John, I did duplicate the same shots I used with the AS100, I just have not included them in my demonstration. Even at 1080, without using Protune, there were far less compression artefacts than with the AS100, and the Medium FOV was far less objectionable in terms of fisheye, particularly when you move the camera. If you look at the demo I have posted, notice that there is minimal distortion in straight objects, like power poles. However shooting in 2.7K and scaling to 1080 in Resolve produces a very much better result.

I will try to put up some shots for comparison.

Geoff
GeeBax wrote on 11/2/2014, 5:55 PM
I have put up two extra clips now, titled 'Tree Test AS100.zip' and 'Tree Test Hero 4.zip', clips 6 & 7 respectively. They are not identical shots, but show the same scenery views, and have been processed through Handbrake using identical settings.

The clip with the AS100 has the SteadyShot feature activated and is shot in Pro mode, while the Hero4 has no stabilisation but is shot in Protune mode. When the shot settles, you can see a noticeable difference in the quality of the image between the two. There was no colour correction applied to the AS100, while the Hero4 only had minimal correction.

Note particularly the rendition of the tree foliage and the grass in the two clips.

www.hollywoodfoundry.com/film%20files.htm

Geoff