Testing transcoding on the Canon EOS 5D Mark III

by planetmitch8 Comments

During our Canon EOS 5D Mark III shoot of the “Incident on Marmont Ave” (to be released soon), we tested the Canon EOS 5D Mark IIIs we had to find out what our best options were for transcoding… our on site media manager and editor Andrew Cohen did these tests and recommended 5DtoRGB by Rarevision.

Canon EOS 5D Mark III transcoding tests



Andrew says:

During production of “The Incident” we were constantly testing the Canon Mark III in all areas to see what changes there were and how to best utilize what it had to offer. In editorial, you want to make sure the video delivered to you is the best it can be, which is why a lot of different software is put through its paces to ensure the highest quality possible. In the case of transcoding, we tested the main Transcoding software on the market and came to the conclusion that 5DtoRGB was the best one when it comes to get the image that closest represents the Native H.264 video recorded from the camera. Where it really excelled was in keeping detail in the shadows/highlights and maintaining a fuzz-free image at all times.

Of course, nothing is perfect and the 5DtoRGB does have a few problem areas, all but quite minor. The first revolves around the fact that within the program you can only have one batch at a time. Thankfully, a quick workaround is simply having multiple instance of the program running at the same time, something that if you have a multi-core machine, should not slow down the computer. The second qualm I found was the length of time it took to transcode each file. On average, transcoding took 2 to 3 times longer than if I had used a comparable program. Again, it is worth the wait as the final video file you get maintains its original attributes without losing anything.

Overall, the 5DtoRGB software really impressed me with its ability to preserve the image without any loss and even slight boosting of the gamma for a more dynamic range.

Disclaimer: these guys are not a planet5D sponsor nor did they give us a free copy – I know it is shocking, but we bought our own! We publish gear news because our users have asked us to let them know when we find things that may help them improve their shooting. If you're a gear maker and would like us to publish your HDSLR related products, please contact planet5D

(cover photo credit: snap from the Rarevision site)

Comments

  1. I can second this. I shot a music video with the 5DmkII just a few days after its official release (still waiting to be allowed to share it online).

    I had never used 5DtoRGB before in the past when working with 5DmkII/7D footage. I decided to buy it from the Mac App Store – and glad I did!

    I compared results converting from the cameras H264 (All-I) files to exactly the same settings with Compressor, MPEGStreamClip and 5DtoRGB.

    5DtoRGB was the only one which left the image looking exactly the same (or at least unnoticeably different). When zoomed in I could see either colour or sharpness differences with Compressor and MPEGStreamclip results… and no changes at all with 5DtoRGB.

    I used ProRes 4444 25fps Rec709 Full range Luminance and left the other settings default. I edited with Proxies in FCP and then replaced the footage after the Grade (done in FCP with FCP tools and various FxFactory tools).

    In my FCP Sequence settings I had ProRes 4444 selected of course, Gamma Correction to Automatic, Render 10-bit material using high-precision YUV selected, and Super-White selected.

    When it came to exporting I rendered in FCP, Exported to QuickTime movie (not QT Conversion) using “current settings”. Then dropped this file into Compressor and exported to an 8000kbps H264, with colour set to “preserve source”… this file seemed to look pretty much exactly the same as the 4444 ProRes (at least in terms of colour)… no shifting.

    So, I was very happy with the results throughout this workflow, everything stayed looking just as I expected.

    Odd that Compressor should preserve colour when going from the ProRes 4444 file to an H264, but shift the colours/gamma when going from the camera’s H264 to ProRes! Don’t get why that is… but 5DtoRGB sorted it ;-)

  2. Not seeing any difference at all when using this vs using Compressor. I’m “only” converting to ProRes 422 HQ (rather than 4444), but I’ve overlaid the same clip done using both apps, and literally can’t find a pixel difference. No visible difference in clipping of highs/lows, no color or brightness shift, no detail difference.

    Wondering if people are doing something wonky in their Compressor presets that would cause some sort of degradation?

    Not seeing the advantage of this tool vs Compressor (unless there’s some sort of difference if you’re processing to 4444, which seems overkill since the source files are 420 encoded H264s…). And with compressor you can set up multiple batches and use multiple processors.

  3. Hmmm. I’m a bit skeptical. I use PPro for all my editing and see no reason for transcoding anymore. Can you elaborate on WHY you feel there is still a need to transcode?

  4. This may be an amatuer question, but why transcode in the first place? I’ve never had any issues editing with the h.264 codec.

  5. The overall goal of transcoding is to maintain the quality of image you shot your film with and keep a full dynamic color spectrum available for when you do color correction and any effects work. When that file is compressed to h.264 or any other compressed format, some of that information is squished into a much smaller package.

    A program like 5DtoRGB changes the file into a format that your preferred editing program will accept and also maintains all original information written into the native file taken from the camera. Essentially you are converting to a accepted format and creating a bit more headroom for color correction.

    With Premiere CS5 and up, when you bring video files into Premiere, it is transcoding your footage in the background so a separate is not needed.

  6. canon fcp7 plugin supports log and transfer to 4:2:2 prores hq, but it appears to only preserve 4:2:0. Going al the way to 4:4:4:4 seems like overkill. Does this tool or any other workflow interpolate back the 4:2:2 from what canon outputs? Canon won’t reply to my support questions regarding this. Guess it is not eager to discuss intentionally downgrading quality to promote the more expensive camera, right?
    Overall I have not seen great video out of the mark iii yet. 60fps 720p 1/125s, kit lens, neutral color mode, etc. tripod with stabilization off. Any other hints for maximum quality?

  7. re’hints for maximum quality?’ sure, exchange your markIII for a mark2 and with the $ saved get a good L lens instead of the kit:)
    But seriously, why are you shooting 720P? What PS are you using? What specifically are the quality issues you refer to?

Leave a Comment