Testing Grinder, EOS Plug-In for FCP, and Compressor
After discovering the issues with the Canon EOS plug in for Final Cut Pro randomly truncating some file conversions I decided to download and test out Magic Bullet Grinder as a potential alternative solution. Then I threw Compressor into the mix as well. Frustratingly, in the end, all 3 solutions had some sort of issue. I've listed my test and the results below. Anyone have any better solution they are using or suggestions on something I might be doing wrong? Are you seeing the same results?
I encoded 4 minutes and 57 seconds worth of 7D footage shot 1280x720. There were 4 separate clips, some featuring bright material and others very dark material. All clips had a lot of detail. In all cases I encoded to the Apple ProRes 422 codec. However, Grinder doesn't provide an option to encode to 720 so I had to up convert to 1080 during the conversion. I'm using a Mac Pro, 2 x 3 Ghz Dual-Core Intel Xeon system with 10Gb of RAM. I read the files off an internal 500Gb 7200rpm SATA drive and wrote them to a second drive with the same specs. I'm using Final Cut Pro 7.0.2, OS 10.6.3. I only have this 1 machine so I wasn't able to see if the problems with each software package carried over to other systems.
Log and Transfer in Final Cut Pro using the latest version of the EOS Plug In from Canon:
Took 20 minutes 10 seconds. Final shots looked almost identical to the originals. However, some shots get truncated. So a 1 minute original shot might end up 20 seconds after conversion with the only solution being to check converted clips against the originals for any differences. A big pain in the behind on shoots with dozens and dozens of clips being converted. From what I've read on the internet this is a common problem and "bug" in the EOS plug in. It kills workflow speed in my opinion as each shot needs to be checked, problem shots re-encoded, checked again, etc.
Magic Bullet Grinder - took 26 minutes and 30 seconds for the same 4 clips. That's a bit longer than I expected since Grinder has the ability to use all your systems cores but that may be explained by the required up conversion to 1080. On the bright side - no truncating issues! But - the converted footage is slightly darker and I think a bit more saturated than the original. This wasn't a big deal on the daytime shots I tested, but some details were definitely lost in the darker night shots. Also, Grinders feature set is VERY basic. I'd love to be able to see more metadata information (ISO, aperture, shutter, etc.) , re-name clips before conversion, preview clips before conversion, mark in and out, set multiple destinations, etc. - none of this is possible in Grinder from what I can tell.
Compressor - took 30 minutes on the nose to convert the same 4 clips. Color wise - they looked identical to the originals. However - all of the clips were cropped around the edges - despite confirmation that crop was set to "none" in Compressor. So it's the slowest solution and knocks off what I would guess is 3-4% of the image around the outside. I ran a couple of extra clips through Compressor just to make sure the cropping wasn't a fluke with the same results.
Anyway - that's my results. I'm not sure yet what I'm going to do for a solution. I've e-mailed the good folks over at Red Giant to see if they have any suggestions for why I'm seeing a color change in Grinder. I'll let you know anything I hear. If it weren't for that - I'd probably stick with Grinder moving forward and may stick with it anyway - only moving select "very dark" shots over to the EOS plug-in for import. When I have dozens of shots to encode every few days letting them go overnight is easy - but having to check them one by one to see if the EOS Plug In truncated any of them, and then re-encode them, is too big of a workflow hassle. I'd rather deal with a slower solution whose results I had stronger confidence in.
Any thoughts or tests you've done yourselves?
I do think the auto upconvert is a bad thing, it is a great feature but I would want to turn it off. If I was using a HVX or HPX I shoot 720p so I would want the Canon at the same resolution. But I would say this is probably why the longer times since it is going more.
You should try the same test with 1080p, my guess is that is what most people are shooting.
The one thing I like about the Plug-in over Grinder is the plugin uses TOD TC vs starting every clip at 0,
I'll try out a few 1080p clips out of curiosity. The clips I used were grabbed from a sports shoot we did a couple of days ago that is going to require significant slow motion speed changes during post production so we opted for a higher frame rate over resolution. I agree with you though, most people are probably working with 1080 material.
Fusion Media Works
Thanks for the post - good to hear some Grinder info. Have you tried NeoScene from Cineform? I've had good results with it thought it's a bit more expensive than Grinder.
I didn't think Neoscene could go to ProRes but I see in the tech specs that it will on FCP systems. Looks like they have a trial version so I'll download it and give it a run. Thanks for pointing it out.
Fusion Media Works
Give MPEG Streamclip a try! It's free, reliable and about twice as fast. BTW, the problem with clips not flagged correctly as progressive can be corrected easily in FCP's browser. Just switch to "none".
Director of the Institute of Media Research (IMF) at Braunschweig University of Arts
I did very similar tests, with similar results.
I noticed the truncated clips problem occurred only if I loaded clips into log and transfer and then kept working whilst clips where still transferring.
With Grinder I also found it created upper field dominant clips, something you don't want in progressive, I could find no way to prevent this. This could also have contributed too, when compared to other transfer methods, problems with more pixelated edges on fine diagonal detail when magnified in the viewer / browser. (This also meant clips had to rendered in the timeline if used in conjunction with other progressive scan clips).
I liked the way the Canon Plugin used time of day info to create timecode.
I use the Canon method with care and check the duration of each transfered clip against the original.
Cinematographer, Steadicam Operator, Final Cut Pro Post Production.
Hi Don and others,
Today, without having read this post, I did almost the same testing. I tried Grinder, Mpegstreamclip, Canon's plugin for fcp and a new one 5DtoRGB (rarevision).
For canon to fcp workflow my choice became the canon plugin. It's a real pitty that the clips sometimes are truncated, cause besides that, it's the most perfect one I think. Almost no visible shifts in saturation and luma and TIMECODE!!! yes! That makes an editor like me happy. It allowes me to work with duplicate frame detection, which I always do. Plus it was pretty fast.
Mpegstreamclip did the baddest job when comparing the files to the original. Too bad, cause I really like the speed and simplicity of this tool.
Grinder also shifted the luma and saturation values.
5DtoRGB did an oustanding job, but has two big cons; no batch converting and no timecode added (or at least not TOD). It also took a lot of time, but for me personally that's not a priority.
Finally compressor. Results were very good, but I also had the cropping problem, cropping the image a little bit. No timecode added.
I hope this contribution has been usefull!
Good luck everyone converting yours...
Avid/FCP Editor, Amsterdam, The Netherlands
Great to hear someone else experienced the same results. I've gone back to using the Canon plug-in too and am just being very careful to watch for truncations. I spent some time going back and forth with Red Giant techs on Grinder and even sent them before and after sample clips. They said they were unable to duplicate the results and it must be my system. When I told them I was getting the same results on multiple systems they stopped responding to my inquiries. So I'm hoping for an eventual update to either Canon's plug in or Grinder that may fix the problems.
If you get a chance - drop the Grinder techs a note and let them know you're having similar issues. That might get them to look more seriously into the issue.
Thanks for the update.
Fusion Media Works
From an overall workflow point of view I think the canon plugin is a timesaver, effortsaver. Also I noticed that it creates the audio as two mono tracks, which can also be usefull. I didn't mention sorenson squeeze in my post, though I did give it a try, but the results were extremely bad looking and not even worth mentioning.
It is still kinda funny that there isn't a single tool that really does the trick like I want it, considering myself as a professional user with very normal demands (eg Timecode).
Instead of writing to Grinder I think I'll give rarevision my opinion on there product cause they really want to make the best tool considering image quality (and not speed).
Avid/FCP Editor, Amsterdam, The Netherlands
Just did a little test myself.
I have a card, where I cannot use the Log & Transfer tool, since the Card was damaged and had to be recovered.
So, I tried MPEG Streamclip, Grinder & Compressor.
Here is my results:
Grinder's Luma and Gamma shift is unacceptable - too dark, too saturated.
MPEG Streamclip is surprisingly good for a free application. Slight shifts, but incredibly fast.
Compressor, looks very similar to Streamclip.
I did not experience the cropping though. The compressor showed the same frame, than MPEG, Grinder or the original file.
I still wanna go for 5DtoRGB, since I heard it is the best by far.
TIME BANDITZ Productions