Personal View site logo
GH2: MJPEG 100Mbps Low Light 1080p Settings
  • UPDATE: This MJPEG patch is compatible with the latest version of PTool 3.65d. The standard non-anamorphic version is also included in Flow Motion v2, which is available for download here:

    http://www.personal-view.com./talks/discussion/3337/gh2-flow-motion-v2.02-720p-reliability-in-auto-exposure-modes/p1

    As numerous GH2 owners have discovered, Vitaliy's recent updates to PTool have enabled the newly hacked GH2 to sustain remarkably high bitrates in MJPEG HD mode. What's even more impressive is that the GH2 maintains error-free recording stability at bitrates significantly exceeding those that can be used reliably on the GH1. And where the GH1 struggled to produce FHD-size videos in MJPEG mode, the GH2 can be readily patched to record in any MJPEG frame size, all the way up to 1920x1080p.

    The MJPEG encoder's most valuable asset has always been its ability to record each video frame as a JPEG-encoded key frame. This enables it to track high-speed action at 30fps without any of the smeared motion artifacts that plague low-bitrate AVCHD videos. To produce high-quality, detailed key frames at 1080p, however, an Intra-frame encoder must work at higher average bitrates than the GH1 was able to sustain. With PTool's support of the GH2, it is now possible to use MJPEG HD mode to produce reliable 1080p videos at an average bitrate of 100Mbps in sharply-focused, highly-detailed scenes.

    The MJPEG encoder also excels in difficult low-light conditions where the AVCHD encoder cuts down its bitrate to minimum quality levels. On the GH1, MJPEG HD mode could be relied on to produce bitrates over 24Mbps even in dimly-lit scenes. With patch settings optimized for consistent bitrate production, the GH2 can maintain high bitrates of up to 100Mbps across the entire exposure range, producing high-quality images under any and all illumination levels. Combined with the GH2's low-noise image sensor, MJPEG HD mode can render clean, gradable images in conditions that the GH1 would have found hopelessly underexposed.

    MJPEG image quality with exposure bracketing

    Below are 100% crops of two series of four bracketed exposures I took in MJPEG HD mode, patched to 1920x1080 resolution. For each shot, I set up the GH2 on a tripod with a fixed aperture at ISO160, and varied the shutter speed from 1/60 to 1/500 second, producing videos bracketed at one-stop exposure steps. Frame grabs were then composited in Photoshop and saved at maximum JPEG image quality. These scenes were selected as typical outdoor subjects shot in natural lighting, as opposed to shots of extremely detailed foliage. The GH2 exhibited no reliability issues whatsoever, and recorded all videos at impressively consistent bitrates of around 90Mbps.

    iPad compatible MJPEG VGA video mode

    In addition to MJPEG HD mode, recent versions of PTool have provided patches for adjusting the frame size and image quality of MJPEG VGA mode as well. Unlike the HD 16:9 aspect ratio of HD mode, VGA video mode records with an SD aspect ratio of 4:3. I've optimized VGA mode to produce 960x720 videos at consistent average bitrates of about 30Mbps. These videos are compatible with the maximum-quality download specifications of the Apple iPad, with its 1024x768 screen resolution.

    GH2 Patch Considerations

    Patches for the GH2 are currently in active development by numerous testers, and Vitaliy has updated PTool several times with exciting new features. The patches in the downloadable PTool INI settings file below have been systematically tested using PTool version 3.61d and have produced no reliability issues to date. This patch is designed and tested only for the fastest Class 10 SDHC cards currently available.

    Additional points of interest:

    * The 100Mbps GH2 Low Light MJPEG 1080p Patch is compatible with all GH2 cameras, and is switchable between NTSC and PAL modes.

    * I recommend using the camera to format your SD card at the start of each shoot, to guard against SD card memory fragmentation.

    * Peak bitrates are obtained only with sharply-focused, highly-detailed subject matter. Average scenes will produce average bitrates.

    * High bitrate MJPEG videos may not play back reliably in-camera or on low-powered computers.

    * If shutter speed is set longer than the frame rate (e.g. slower than 1/30 at 30p), low-quality video files may be produced.

    *** WARNING ***

    The 100Mbps GH2 Low Light MJPEG 1080p Patch will not work properly with previous versions of PTool. My thanks once again go to Vitaliy Kiselev, for his pioneering work on PTool, and to cbrandin for his invaluable Stream Parser tool.

    The best way to insure that you have the latest PTool release is to download it directly from the following link:

    Download PTool here: https://www.personal-view.com/faqs/gh2-hack/ptool-software

    100Mbps GH2 Low Light MJPEG 1080p Patch Settings File:

    For use with PTool v3.62d and above, I've prepared anamorphic as well as standard aspect ratio settings files:

    * For use with standard lenses, the standard patch will produce two types of MJPEG 30p videos:

    HD mode: 1920x1080 HD videos, with peak bitrates up to 100Mbps.
    VGA mode: 960x720 iPad-compatible videos, with peak bitrates up to 30Mbps.

    * For use with anamorphic lenses, the anamorphic patch supports two different squeeze ratios:

    HD mode: 1920x810 videos for 1.33X anamorphic adapters, with peak bitrates of 100Mbps.
    VGA mode: 2160x810 videos for 2X anamorphic adapters, with peak bitrates of 100Mbps.

    These zipped INI files can be used to apply complete patch settings to firmware loaded into PTool. To use, unzip the INI file into the same folder as the PTool application. Launch PTool and load the firmware for GH2 v1.0. The settings contained in the INI file will automatically be installed in the "C" button at the bottom of the PTool main window.

    Note that with the following PTool Settings files, you may at any time re-install the original Panasonic GH2 v1.0 firmware into the camera. You may also copy each type of patched or original firmware to separate SD cards, and use them to quickly switch between patches as often as you like.
  • 112 Replies sorted by
  • Thanks so much lpowell! So glad to see your contribution to the GH2. At the bottom of your post above, there are some GH1 references that I think should be GH2 (Copy/Paste?). Of course, your contribution would not be possible without Vitaily's PTOOL and cbrandin's, and so many others who contribute to this great effort.
  • Thank you lpowell,

    which sd card is recommended? I´m using a Transcend Class 10 32GB Card and sometimes there are card writing errors.

    I wonder why 2 same clips filmed in succession have different bitrates with the same settings. One clip has 90 mbit, the other has 38 mbit.

  • does this patch apply only to 1080 30p, or can you use this for 108024p and 108060p as well?

    also, what kind of timeline or settings would these types of files be edited in for adobe premiere pro cs4 or 5

    side note: I would like to eventually use these patches in etc mode. will that be possible?
  • Thanks LPowell. Reliable high bitrate MJPEG sounds great.

    @Vitaliy MJPEG preset tied to PTool? Are you sure it's good idea?
  • @peterosinski
    This is MJPEG only patch, do not affect AVCHD in any way.

    @MJPEG preset tied to PTool? Are you sure it's good idea?
    Yes it is good idea. As having bunch of reliable, very reliable and low light reliable patches confuse people.
  • I just tried these 1080p MJPEG settings...at the same time dropped AVCHD to 35 H and 30 L presets. Did a very quick test while the sun was out...

    I do like the clips using these MJPEG settings...averaged 102 mbps (using 14mm pancake). I was never impressed by MJPEG on the GH13...period.

    The few A vs B comparisons I just did with same subject AVCHD @ 28 mbps have impressed me as to what MJPEG can now look like. AVCHD still had slight advantage with less "artefacting" in straight lines (e.g. car door edges and window edges) than the MJPEG...but the fine detail resolution was almost indistinguishable from the AVCHD. I like it enough to do more careful testing. Thanks IPowell. Now MJPEG is an option for me.

    Also, even at the reduced AVCHD settings there was still a couple of seconds of corruption at the start of each clip (but the MJPEG encoder does not have this problem even at 100 mbps)...
  • GH1 MJPEG delivered 720 "real" resolution at most and setting it to 1080 would just interpolate extra pixels in camera.

    Is GH2 behaving the same way? Is the actual image 720p extrapolated into 1080p and then encoded into MJPEG?

    Regards,
  • I can't say for sure exactly what's going on regarding 1080p with these particular settings (and I wondered this myself)...BUT it does look very good. Much, much better than anything (and I tried several different MJPEG settings 720 and 1080) I ever managed to get out of my GH13's.

    As I said, apart from some virtually insignificant artifacts which seem to be a characteristic of MJPEG, the image at about 100 mbps is really lovely. This has made my day ...and makes my GH2 worthwhile even if the AVCHD start-up issue isn't fixed.

    The comparisons I did this afternoon with AVCHD strongly suggests to me that something better is happening that 720p being scaled up to 1080p. Even if it is, the quality of the interpolation is excellent (considering that even at 100 mbps the 1080p image quality on the GH13 was mediocre.

    More good news... testing with some of my other wide angles didn't once get a lock-up ...even on my "bush-of-death" (Sandisc class 10).

    I forgot to mention that I varied the settings slightly by NOT ticking anything to do with 480p. Shouldn't make any difference...just for the record.
  • @IPowell
    the 1080p MJPeg is awesome: thanks!
  • On vimeo.com/26836032 I see some blips in the center @ the 1 second mark. Is this normal?
  • This all sounds great. But will the MJPEG patch provide a noticeably better image in some circumstances than just my vanilla UNhacked GH2? Seems to be conflicting info and opinions and a dearth of AB tests. Some reports are that the hack produces more noise -- but allegedly happy noise! I see a lot of nice hacked footage out there but it doesn't tell us much unless you can see what an unhacked gh2 would do in the same conditions. If anyone has a link to a good AB shootout with a hacked and unhacked GH2, please link me. I have no interest in anything that compares it to GH1 or GH13.
  • @brianluce
    why make do not an AB comparison test?

  • @brianluce

    I have done AB comparisons under controlled circumstances. Higher bitrates definitely improve temporal behavior in a measurable way. Higher bitrates also improve detail because lower quantization values can be used (also measurable). We haven't figured out how to get even lower quantization values in 24H yet, but that will come.

    The way this particular codec works is that it estimates how detailed macroblocks should be allowed to be. Sometimes it gets the estimate wrong. The result is that lower in the frame you will see mud or stuttering motion because the codec had to abandon additional processing because it ran out of bandwidth.

    This is all measurable. Furthermore, I have seen clips posted that showed all this.

    As for subjective analysis. Well... subjective analysis can always add to technical analysis. Ultimately, though, you'll never "see" anything a codec is not capable of producing. Nothing gets fixed without some sort of controlled technical analysis. Also I find it unlikely that something that is measurable and repeatable won't be visible - it's just a lot harder to find strictly visually, especially if you are looking for something that is missing rather then something that is there. I've seen lots of footage that "felt" better or worse, although it was not obvious why. Controlled technical analysis usually got to the issue. Note that I am not talking about pixel-peeping, that's still pretty subjective.

    Chris
  • Before dismissing this MJPEG patch, presumably based on what could be achieved with the GH1, it might be prudent for people to try it for themselves.

    It may be an entirely subjective view, but from what I've seen so far GH2 MJPEG clips treated respectfully with a good decoder are not too far off their AVCHD counterparts. It's no reason to switch over from AVCHD, but it remains an interesting development given that I didn't care much for the quality of GH1 MJPEG clips.

    I did some tests earlier, and in the majority of the clips the MJPEG sharpness is almost toe-to-toe with that of the AVCHD clips. Add some light sharpening in Avisynth (with LimitedSharpenFaster) and the two are more or less level.

    @EOSHD You cannot expect to introduce yourself to someone's topic in such a brusque manner, attempting to discredit their work, then complain when they take it badly. If anything, LPowell's reply is somewhat restrained.
  • @cbr, thanks for the reply, did you by chance post any of your tests on vimeo or some such place?
    I know it's early in the process, and there's a lot of excitement and maybe it's too early for shootout type footage posts.

    @butt
    >>>why make do not an AB comparison test?>>>>

    Not sure what you mean here. If you mean why make an AB test, I'd like to see how much difference the hack makes over a stock GH2. So far it sounds like the hack provides a clean motion signature but introduces more grain.

  • There isn't really anything to post. The test shots are uninteresting except that they can be used to feed image testing tools. I typically post results at the end of a test. It's all pretty dull technical stuff, though; mostly useful for people like Vitaliy and LPowell.

    Chris
  • If you are using high ISO values the hacked values will show more noise (actually, they will show more of everything). Good noise reduction plug-ins will be able to differentiate between increased noise and increased detail much better than the camera, though, so the end result (especially with good post processing) will be better.

    Chris
  • Is possible to use this patch with 720p too ?
    I don't like upscaled 1080p, 'cose i see more noise at 3200 ISO
  • Yes, you can set the MJPEG HD width and height patches to any dimensions you prefer, up to a maximum of 1920 wide and 1080 high. For best results, use dimensions that are evenly divisible by 2, and with an aspect ratio of 16:9.
  • I'm too busy making a living out of cinematography to waste my time on a pissing contest. But really...WTF...to imply that highly experienced, working photographers don't have the ability to do their own tests and decide what settings/codec is best for their application. That's a bit rich.

    Besides...I haven't even bothered to do a "low light" test yet....

    If I try a new MJPEG patch on a new camera and it works better than the crappy MJPEG did on the GH13...all I need to say is thanks to the person (IPowell) who brought it to my attention. If you don't like it ...don't fucking well use it...

    By the way, I read EOSHD most days...
  • I would say that any settings that work and deliver something stable / production worthy are a benefit to the community. The more the better and be it Low GOP or MJPEG 100mbit it all comes to choice in the end.

    It is of no ones business to pitch one over the other. Every research gets us closer to a better goal and not to individual fame. I believe credit should be given to every patcher where its due but the kind of kid brawl here is uncalled for.

    Please lets get back to testing all of these great settings and leave the fights out of this forum. Thank you!
  • @EOSHD
    As far as I remember I told you to stay out of this topic.
    And it looks like you have problems with reading :-)

    Topic cleared and some content with value moved to new topic:
    http://www.personal-view.com/talks/discussion/492/eoshd-crusade-on-mjpeg-hacks
  • Confucius say, "Man who complain about SEEING noise, not using his senses and sensors properly, but will be HEARING from the censors soon."
  • if i had avchd 24p clips in my timeline ,how much difference would that be when merge with this new 100mbps mpeg 30p clips after render