HDSLR Encoding Wars – Premiere Pro vs Final Cut Pro
Browsing the web reading various reports and reviews about Premiere Pro I keep coming across the statement that because Premiere Pro works without the need to transcode H.264 DSLR files the footage inherently retains more quality.
After using Premiere Pro for a week or so I decided to spend some time investigating this theory and came up with some interesting results that I thought I’d share with you. Firstly let me describe the workflow involved in each process.
For both workflows I used the same H.264 mov file from an interview I shot recently on the Canon 5D mark II. The file has a runtime of 4:41 and in it’s original state is 1.5GB. This was a particularly tricky shot for the codec as the background contains shaded solids caused by a natural vignette from my 70-200 lens. The H.264 codec in the 5D struggled with this and the source footage contains some macro blocking, but nothing too terrible. Lets see how the two workflows deal with it…
Throughout this post you need to have your browser full screen and click on the images to see the detail, the small ones below don’t really show much difference.
Original H.264 file from the Canon 5D Mark II
Final Cut Pro
• The h.264 file was imported using Canons E1 plugin with the transcode set to ProRes 422(LT)
• Once transcoded / imported the clip was placed on a sequence with the same settings
• The sequence was exported using the ‘Quicktime Export’ option in Final cut with ‘make self contained’ enabled
• The result was a 2.67GB ProRes file
Here’s the same frame after that process. The result looks pretty much the same as the original and of course retains the macro blocking recorded by the 5D when shot.
FCP Quicktime Export as ProRes 422(LT)
Next the ProRes file above was opened in Compressor and encoded back to a 1280×720 H.264 quicktime file. I used this as an example because it’s the setting I use for uploading to vimeo and I’m used to the results that are normally achieved. I set the bit-rate to 5000kbps with a forced keyframe every 25 frames.
The result was a 191.6 MB file. Note the worsened macro blocking in the blue/grey wall on the right, this is caused by the encoder struggling with the natural vignette caused by my 70-200 lens.
1280×720 H.264 Encode @ 5000Kbps from Compressor
Premiere Pro
Premiere Required a lot less steps because it works directly with the source footage.
• The h.264 file was imported into Premiere and didn’t require any transcoding
Here’s how the frame in question looks in Premiere itself, this is a framegrab taken from the source monitor and you can see that the full quality from the original mov file is retained and to my eye does indeed look slightly better than the file transcoded to ProRes by FCP.
Premiere Pro Source Monitor
The clip was then placed on a DSLR 1080/25 sequence and then exported as a 1280×720 H.264 video using a max bitrate of 5000kbps and a keyframe every 25 frames.
The result was a 181 MB file. Unfortunately the gain in quality is all lost by this point and the macro blocking in the blue/grey area is a lot more prominent than the FCP/Compressor export at similar settings. The white wall to the left of the image looks a lot worse in this one with bad macro blocking throughout. Note also that the skin tones on the subject look softer / less detailed than the version from Final Cut as well.
1280×720 H.264 Encode @ 5000Kbps from Premiere Pro
It seems that even though the footage within Premiere itself is holding up really well it’s degrading more upon export than my FCP workflow. I spent many hours testing lots of different export options, codecs and formats as well as playing with options in the preferences, I still couldn’t export a file that matched or exceeded the image quality in the file exported from FCP.
What is the magic setting to export from premiere and retain that quality?
As I stated at the start of this post the original footage from the 5D was far from perfect, it already contained compression artefacts and as such does not make for a perfect test. That being said this makes it a great real world test as after all these cameras don’t shoot perfect footage so we have to deal with the nasty h.264 codec thats being produced one way or another.
I’ll finish off by reinforcing the fact that I’m very new to Premiere Pro, this could very well be due to user error on my part so if there’s something obvious I’m missing or any other tip you can give please don’t hesitate to leave a comment.
Let me know your thoughts.
Paul.
i want to know encoding 5dm2 file to dvd wirh good quality
.which software should i use?
Great info here. I just purchased a Canon 5D Mark II and am glad there are favorable comments here about PPro CS5.5 and Cineform Neoscene which happens to be what I’m working with.
One thing comes to mind regarding the Adobe media encoder settings. Did you have the “use maximum render quality” option checked?
It does have a significant effect on the resulting footage especially when scaling down from timeline/original footage size.
I also have a canon XF 300 that cannot be used with cs4 premier pro… be forewarned Adobe will make you upgrade. If anyone has found a way around this let me know. I want to be able to edit those MXF files with Premier Pro CS4… please dont force me to upgrade to CS5, Im just not ready!
I use the same set up! the 5d MII with Premier Pro. My workflow is simple and effective and looks wonderful! You can check out some of the videos I have shot with my 5D on my You Tube page under the same name, Bourgeois Photography.
Cheers! Thanks for writing this! Great info!
Bill from Adobe here again. Wanted to follow up on my note from December on this topic.
Well – our team has gone back and done the homework and we can now say categorically that we are correctly reading every pixel as it is captured and presenting it without alteration. There is a lot in that statement as I know as we’re talking about Color Space conversion from YCbCr to RGB as well as decompression. On a Mac, you can lean on QuickTime to do this – or go get the pixels yourself and have full control. We do the later and this is the root of what is different in the results out of our respective products.
As we’re dealing with an image that has been scaled and compressed, our philosophy is to preserve every bit of color, light and detail we can – so we apply no filtering, as such, when you do this with some images where the there are broad areas of the image have gradations of similar color – and you apply no filtering, you get the results we saw on this thread. However we found that if you introduce grain at the front end of the decode chain you end up with a more appealing image in these cases, however, any noise introduced reduces picture quality – not always noticeable when you’re looking at a small 320×240 image on a web site, but may be of concern for a high resolution workflow.
We’re not alone in our findings – we found some other folks who apply the same philosophy as Adobe, Rarevision has a stand-alone tool that follows the same processing as the Adobe Pipeline – you can see some of their findings on their website (http://rarevision.com/5dtorgb/). If you test the approach that Adobe and Rarevision use vs. QuickTime and get into the details – you’ll see some differences, particularly in transitions between bright and dark areas and color edges; the net – you end up with a higher quality image by getting the pixels natively.
We’ve also done some benchmarking on the final stage of encoding for distribution – we’re happy that we can produce the best looking results, faster than any other solution – but I’d encourage you to test in your pipeline and let Adobe know if this is not the case for you.
So – there you have it – we give you the pixels in the purest representation as captured – what you do with them from there is up to you. QuickTime? Well you’d have to ask Apple, but it seems that there is some form of image alteration being applied, but we can only judge by the results.
Hopefully this doesn’t stir up a hornets nest – but our intention is solely to state that the results you see out of an Adobe pipeline are the purest representation of the assets captured in your Canon camera that we can possibly present.
Paul – thanks for putting this out there for debate – your posting did encourage us to go back and check the entire pipeline and we’re happy we did.
I’ve never worked with DSLR footages in FCP or PPro. But I always edit in boths apps using ProRes HQ. I’m working as a MoGraph artist in the Broadcast industry and H264 is one of the materials we receive and edit(we handle mixed formats a lot). I have set my FCP to output all the time a High Quality 10bit proress HQ by changing some preferences on the Sequence and Render process in FCP. We always output in ProRes HQ and do the convertions out from the exported files… H264 etc…
I’m now editing in PPro and still working on ProRes. I set my ProRes Preset in Media Encoder to 48bit/10bit maximum depth render. I find PPro having better ProRes HQ exports. Encoding the exports to H264 in compressor or MediaEncoder has the same results…
On a MAC my workflow has always been…
Orginal Codec -> ProRes HQ -> Edit -> Export ProRes HQ -> Other Formats.
I forgot… Make sure FCP is set to render and edit in Full High Quality 10Bit rendering… In PPro create a custom(Desktop) sequence as ProRes HQ and make sure during export always set to “Render Maximum depth” and 48Bit encoding.
I hope this info can help you guys…
Thanks for your feedback Wilbert, that sound like very valuable info.
Best regards
Paul.
I am using premiere cs5, importing mpeg file everything was ok, but next day i try to import again a mpeg file from a canon and only import the sound, I am new on this. I did something wrong??? please somebody can help me.
And thanks Paul this is a great place
“The sequence was exported using the ‘Quicktime Export’ option in Final cut with ‘make self contained’ enabled
The result was a 2.67GB ProRes file…
Next the ProRes file above was opened in Compressor and encoded back to a 1280×720 H.264 quicktime file…”
The discrepancies you’re seeing are likely the result of using Quicktime to do your H.264 encoding. For more background on this issue, do a search on “Quicktime gamma bug”.
If that is the cause, then it’s not a problem that Adobe can correct. (But you could work around it by installing a third-party H.264 codec.) Issues with H.264 conversions were resolved in the Win7 version of Adobe Premiere with CS5’s codec plug-in architecture’s ability to bypass Quicktime.
Thanks for the suggestion but the file exported using quicktime was the better of the two. I tried many different export codecs from premiere and the problem persisted. In the examples shown actually using a MainConcept H.264 encoder.
Following this thread, I have done extensive tests with
Adobe CS5 Media Encoder and its transcoding of AVCHD to ProRes422
standard vs. Final Cut Pro’s Log & Transfer encoding of
AVCHD to the same ProRes format. Every single time…the ProRes
that was encoded by Adobe Media Encoder was SOFTER with CHROMA
BLEED on sharp edges (seen under magnification). I posted the test
results with photos on the Adobe Forum, but have never received any
acknowledgement of this issue from Adobe…even after having
emailed their tech support people. I have wanted to use Adobe
Premiere Pro, but because I do the bulk of my editing in Pro Res, I
have stayed with Final Cut Pro for the sharper image without the
chroma bleed. The test results are summarized in this link to the
Adobe Forum: http://forums.adobe.com/message/3073040#3073040 If you
check out the photos, I believe they really illuminate the issue
and make it easier to understand what is going on. If Adobe is
working to correct this, I hope this link will be of help to them.
Thanks, John
Firstly, great write up! Secondly, wow doesn’t the
compressed Premiere export look absolutely shocking… just as i
was about to convert… Hmmmmmm i hope this gets sorted out, was
nearly sold on CS5!
To be fair the general encode quality from Premiere is fine, this specific clip however is highlighting a difference in the way it handles noise in large areas with dark graduations and for some reason compressor handles that slightly better.
I’ve been using Premiere on all my jobs lately and have been happy with other projects that are either better lit or show less dark graduations.
Paul, Any update or further testing of this issue?
Not as yet. I’ve been using PP and Media encoder a lot recently though and have been getting good results. I guess that shot just found it’s weak spot.
Paul,
I’m responsible for Product Management across the video & audio solutions that Adobe provides.
Firstly, thanks for being open and fair in your comparison between the workflows, as a result of your efforts our teams are already working to understand why you would experience differences in the end result.
As you say – the footage in this workflow is challenging and not representative of ‘typical’ video, which may account for our some of the differences; but no excuses – we want our workflows to provide the best results in the shortest time possible with ANY content. We’ve already found that our default presets for some web streaming formats are 50% lower in bit rate than other products and are set to do less preprocessing (to get a faster result), but I don’t believe this is what you came across in your tests, so we’re reviewing the underlying technologies, if an issue is uncovered, rest assured it will be fixed.
Note that we may be slightly slower in our response due to some of our folks being out for the holidays, but we’ll post again to your blog when we’ve got news on this topic.
Thanks,
Bill
Hi Bill.
Thanks very much for taking the time to leave feedback, I’ve been really impressed by the way Adobe representatives have been getting involved and answering my questions since starting out with Premiere Pro. I’ve been trying various export techniques and workflows all week and also communicating with Wes Howell on the subject who has also been extremely helpful and a pleasure to work with. Since writing this post I’ve come to realise just how much trouble this shot is, it seems to be the combination of compression noise and graduated colours that’s proving to be so troublesome and is probably any encoders worst nightmare!
It’s great to know you guys are on the case though and pushing forwards. I have to say that over the last two weeks that’s the one thing that’s stood out in comparison to working with Final Cut Pro, it really does feel like the adobe team want to make the product be the best it can be and not only reply to questions but actually want to contact users and work with them to solve problems and better the product.
Best Regards
Paul.
I’ve been trawling through the internet for most of this year trying to find out if there is a significant difference in capture quality between Final Cut Express and Adobe Premiere CS5. I use a Canon XHA1 (tape). When I capture with FCE it uses the ‘Apple Intermediate Codec’, which by the time it gets to DVD the whole production looks soft and a little vague. I understand that I would need a Mac Pro and Final Cut Pro (etc etc) in order to use the ProRes codec.
So I tried a tedious ‘live chat’ conversation with an Adobe representative to try and find out, would Adobe’s capturing process result in better quality. Overall they said ‘yes’ but didn’t provide a reason, instead making constant reference to their inane PDF documents, (which also didn’t provide anything awe inspiring.) Could anyone enlighten me, before I commit over AU$500 on this video editing software? At present I’m using a MacBook (2.4 GHz Intel Core 2 Duo, 4GB 667 MHz DDR2 SDRAM)
Hi Peter.
Although I can’t actually answer your question specifically but I do know that encoding HDV or most HD formats to DVD can often lead to soft unimpressive results. Certainly encoding with something like Apple Compressor or Adobe Media Encoder would give you more options to adjust encoder settings but it’s by no means a guarantee that you’ll then get better results.
Personally I doubt it’s the capture that’s the problem, it’s more likely to be the encode stage.
Peter,
I have to agree with Paul. Your problem is not in the capture or at least shouldn’t be. I haven’t used Final Cut Express much, but I was wondering why you capture it in the Apple Intermediate Codec instead of just capturing it in the native HDV format, I know it is really compressed but it actually holds up relatively well. If you are worried then you should transcode it after you have captured it (takes a little more space but could solve some problems) FCP likes to have firewire based tape formats captured as they were shot, the only way I would transcode anything tape based while capturing was if you have a capture card like Blackmagic, AJA or Matrox so that you can capture and convert at very high quality.
Also you do not need a Mac Pro to use ProRes, and iMac will run it and so will all the new MacBooks. When it comes to using a codec like Pro Res, it is more about how fast your hard drives are. RAIDs are great.
I feel as though I was living my life in a constant state of frustration when creating my files for DVDs, Vimeo, and most anything that needed compression. Although I am a staunch advocate of FCP, I’m not a fan of Compressor at all. About a year ago, I tried Sorenson Squeeze and my days (and nights) of frustration are over. Squeeze has many presets and just does the job right – all the time. They have a free download in case you’re interested.
I did try Premier Pro and as much as I wanted to love it, I just couldn’t. I do use Encore because it flat out does a much better job than DVD SP.
I use the same workflow as Paulie but i’m still cutting with FCP (I also tried to work with Premiere) but I’l waiting what’s happen with FCP.
Steve Jobs promisse a big Update at the beginning of 2011.
Cineform’s NeoScene is was Paulie describe. I compare the output from NeoScene with MPEG Streamclip, Canon FCP plugin and other’s but NeoScene give me the best results.
Sorry for my english.
Hi Paul
Here is a link to the presets which I use:
http://www.precomposed.com/hd2sd/ame_hd_mpeg2i_cs4.zip
essentially these are MPEG 4:2:2 at 100Mb and are huge files but the quality is extremely good.
When Transcoding for DVD I use HC Encoder, AVISYNTH and hd2sd, these are all available free but only on the PC and I’m assuming you’re a mac guy. The DVD compliant MPEG files produced are very very good.
You may want to take a look at this discussion which I found very useful when researching my own workflow.
http://www.precomposed.com/blog/2009/07/hd-to-sd-dvd-best-methods/
by the way if you do wedding videos (I see your partner does wedding photography) Jon Gedde’s site precomposed.com has some fantastic 3D DVD menus for weddings.
Thanks David, very much appreciated. I’ll give that a try.
Let me know how you get on
Sorry I’m not sure if you shot this at full 1080 – assuming you did I suspect the problem with Premiere is during the downscaling part of the transcoding.
Premiere isn’t partuclarly good at this. When downscaling I always export out of PPro @ MPEG 100Mb Iframe 4:2:2: and then use an external encoder to do the downscaling. This workflow introduces an extra step but the gain in quality is very significant.
If using PPro to do scaling them make sure Frame Blending is off and MRQ (Maximum Render Quality) is on – this puts it into 32bit color space.
Additionally ensuring you use the CUDA technology to transcode will increase the quality as different scaling algorhythms are used.
More details for those interested can be found here:
http://provideocoalition.com/index.php/adobe/story/scaling_in_adobe_premiere_pro_cs5/
Concerning CS5 and the MPE I have to say it is quite incredible since moving over to it my productivity has at least doubled and it allows me to be much more creative by enabling more complex editing.
Thanks David.
I’m trying to export out without downscaling and still losing a lot of quality. Would you mind describing a bit more about the details you’re using to export?
For example, which format, which MPEG type etc? A screengrab of the export options page would be terrific if you could?
Many thanks
Paul.
I still transcode my 5D Mark II footage with Cineform’s NeoScene before editing in Pr CS5. It expands the native 5D2 file into a 4:2:2 10-bit colorspace which allows more headroom for grading and encoding out to other formats. And, the quality is honestly better. Adobe’s effort is appreciated, but they aren’t there yet. The NLE is great, couldn’t be happier, as long as I’m feeding it 4:2:2 .AVIs from NeoScene. The MPE works VERY well with the transcoded files. Only one drawback: converted files are 3x bigger.
So now that I have NeoScene, the Adobe integration with AE, Photoshop Extended and Soundbooth make it the no-brainer editing suite. Plus, the 5D2 is an amazing stills camera and the Adobe suite offers Photoshop. What do you get with FCS?
;-)
Fully agreed. Premiere & AE + CineForm is easily the best quality, most efficient choice.
That’s interesting, thanks for that guys.
The main appeal for me with Premiere over FCP is the workflow without pre-transcoding and it’s ability to handle effects etc without rendering with the MPE. If it turns out that I need to pre-transcode to get good results I’m not sure I’d be so keen to re-learn and move away from FCP.
Are you serious?
Why would you use ProRes LT? Should I link you the Wikipedia article on ProRes so you’ll know what the LT variant is for? I absolutely do not believe that using the normal or HQ variants would produce any more macro blocking. It never has in my experience, their data rates are more than twice what you shot at.
Not sure you understand what “native” means. It just means that Premiere is building I frames on the fly instead of all at once.
Going over screen shots with a fine took comb when you’re shooting to a compressed format in the first place is silly. If you’d like to do any sort of testing that means anything feel free to use a chart next time.
I use ProRes(LT) because as you say it has more than enough headroom for use as an H.264 intermediate. Apple’s white paper on ProRes also states that the (LT) variant makes an ideal intermediate for use with lower bit-rate complex codecs.
Regarding the chart, yes it would probably be more scientific to make an encode using a chart but I’m just sharing real world results and problems here.
Interesting stuff, it kind of contradicts what Shane Hurlburt has been saying about the native codec providing incredible results in CS5.
I’m still using FCP and I’d be forgiven for thinking that CS5 is a bit of a one trick pony : you can work natively with DSLR H264 files (and this is the major selling point of every piece of ad blurb I’ve seen for CS5). What about all the other codecs and cameras and 1001 other reasons why FCP works so well. Does CS5 address them?
In any case, the h264 advantage largely disappears the moment you start adopting a more robust and ‘professional’ attitude to what goes into your project and what stays out.
Log your footage, only ingest what you think you’ll need (with subsequent transcode to ProRes) and the benefits of working with native h264 in CS5 suddenly appear much more marginal.
In a workflow where you have to consider grading, metadata, masters, storage, archive and future retrieval, the benefits being able to lob a bunch of h264 MOVs into a bin and go straight to edit is a relatively minor issue that throws as many complications as it does advantages.
Thanks for your comment Terry, some very good points there. The Shane Hurlbut point is an interesting one because while the footage is in Premiere it does indeed hold up well. The problem is how do you get it out and retain that quality?
Regarding the workflow, I guess it depends what you do. Personally I’m benefiting greatly from editing in Premiere at the moment. This is probably because I’m working on projects that benefit from fast ingesting without any logging etc. The workflow is more efficient for me and I end up with an archiving system that makes a lot more sense than having multiple copies of masters in different codecs to deal with.
It actually doesn’t contradict what Shane Hurlburt has been saying because he does not edit with the camera native file, he first transcodes the h.264 file to ProRes and then edits it. The process he uses is exactly the same as the process for editing it in FCP. He said that by doing it in Premiere the colors stay closer to the original, and if you look at it on a properly calibrated broadcast monitor it does. The files created by FCP add contrast making the darks darker where Premiere does not.
I am not sure why everyone wants to edit these files in their native format since it is far inferior to ProRes and makes the computer work much harder.
One thing from this article that I do not understand is why in FCP he transcoded the files to ProRes then put them on the timeline and exported a ProRes files from that. It is just one extra step that is not needed because when you transcode the files to ProRes it creates the ProRes file already and puts it where ever you have the files being captured to. Just use that file and save yourself from duplicating all of your footage and compressing it and extra time.
Saying that CS5 does not transcode the native is a bit incorrect. In fact CS5 uses the Mercury engine to transcode, on the fly, for any frames that need it. For example, if you apply some color correction, CS5 will transcode the appropriate frames to a 4:4:4 colorspace.
Not sure this helps to understand why the results differ but it could be something to add into the equation.
I make a kind of similar test; in FCP i do a Log and Transfer then to Compressor and generate a ProRes file. In Premiere Pro just load the file and use Media Encoder to export as ProRes file.
Well the quality of the Compressor ProRes was a lot better than the Media Encoder, retaining more detail and contrast and the Media Encoder was a little washed and soft.
Clip file used was a AVCHD from a GH13, 1920x1080i converted to Progresive.
Software used: Final Cut Studio 3 and Adobe Creative Suite 5
The sad thing is, i´m a Premiere user ;)
Since I am not able to use Premiere because It’s really too slow and lagy on my Mac I ran on the same quality problem. I realize that it was compressor that gave me really bad result.
I realize that when I made my first edit with premiere with an old clip and compare it together.
Then I drag the clip into compressor to have the same setting and again huge difference.
So What I do right now is I use media encoder to compress my FCP pro Res montage at the end. Another reason why I will stick to FCP is Apple Color.
Thanks for your test I really wish Apple will edit those file without transcoding