GH1 Action Short / Field Test

[John Caballero]

The AVC support from Panasonic is originally for the HMC 150. It should work with the GH1 1080p. I am in the process of experimenting with that workflow. I am using some original files posted here. But I really going to get more into it when I get my camera. Its in the best interst of Panasonic to make sure they support their AVC formats as they do P2 MXF.

 

[PappasArts]

We reshot with a lower ISO yesterday and no noise, almost too clean. .

No such thing as to clean............ I would rather start out noiseless ( on the senor level ) then add through ISO or post to dirty it up. To have the choice is better then to not to have one; every and any time!

Still confused why your footage ( large downloaded file ) looks more noisy, videoish than ( comminderspike' ) night footage shot at 400-800-1600 iso in 1080/24 mode. Blooms desert piece as well looks more filmic and that was in 1080 mode. Khoile GF shopping spree vid and hunters stills also look more filmic. I think your having some kind of pulldown artifacts too unrelated to the above obsevations...

 

[DrBlaz]

Still confused why your footage ( large downloaded file ) looks more noisy, videoish than ( comminderspike' ) night footage shot at 400-800-1600 iso in 1080/24 mode. Blooms desert piece as well looks more filmic and that was in 1080 mode. Khoile GF shopping spree vid and hunters stills also look more filmic. I think your having some kind of pulldown artifacts too unrelated to the above obsevations...

same here! also the 720p60 parts use 1/50 so its really 720p30, exactly like mjpeg parts, the stuttering seems to be related to 30p-24p conversion.

better use a HDV at 24p for these parts... more quality and the DOF is the same.

 

[John Caballero]

also the 720p60 parts use 1/50 so its really 720p30, exactly like mjpeg parts, the stuttering seems to be related to 30p-24p conversion.

better use a HDV at 24p for these parts... more quality and the DOF is the same.

He clearly explained what happened with the 1/50. HDV? LOL.

 

[PappasArts]

He clearly explained what happened with the 1/50. HDV? LOL.

However,

What hasn't been explained is why the footage ( large downloaded file ) looks more noisy, videoish then ( comminderspike' ) night footage shot at 400-800-1600 iso in 1080/24 mode and exhibits less noise. As well, Blooms desert Joshua Tree piece looks more filmic, and that was in 1080 mode. Kholi's girlfriend running a muck in a thrift store shopping spree vid and hunters stills also look more filmic and less noise. Pulldown artifacts from the extracted 1080/24 are apparent as well! Why would this be? I thought the process for perfect extraction had been addressed too.. This camera has done better; why in this test has it's performance seem to take a hit compared to others mentioned above....





.

 

[Isaac_Brody]

However,

What hasn't been explained is why the footage ( large downloaded file ) looks more noisy, videoish then ( comminderspike' ) night footage shot at 400-800-1600 iso in 1080/24 mode and exhibits less noise.

My guess is encoding compression and the fact that most of the footage was shot at 30P and 60P.

 

[Noel Evans]

Great Jack. Thanks so much for posting all this info. Sometimes I love being in PAL land without all the pulldown hassles.

 

[Isaac_Brody]

One more thing, the 30P MJPEG footage I've seen looks like it has some sharpening applied at default settings. I've seen this on some other 30P shots from the GH1. I'm curious to see MJPEG with sharpening dialed all the way back.

 

[Jack Daniel Stanley]

Thanks Jack, for posting this along with the RAW and settings. I also noticed the stuttering as well, and I believe it to be an artifact of the 30P to 24P conversion (which by its very nature is stuttery due to dropped frames). And this is what concerns me a bit as it being used as a valid workaround. So many people have been saying that they'll just shoot in 720P modes and convert, but it doesn't always look right. As we've seen here it can actually be distracting.

I don't know about distracting - if we're seeing the same thing anyway - because I've watched it over and over again and I'm only 80% sure I even know what you guys are talking about. Barry Green says he's not seeing it in the 200mb file either.

But I kind of think I'm seeing it now and am researching the 30p tp 24p thing and will try a new method and detail the new and old methods shortly.

It's tricky and it will work in a pinch, but to rely on it as workaround seems a bit risky in my opinion. If Panasonic had just given us a 720-24P MJPEG mode so much of this could have been avoided. It's nice work though, Jack, especially considering how quickly you turned it around!

Although I didn't do it by dropping frames. Which is what it would be if you just dropped it into a 24p timeline. I did it by interpolating to 24p via compressor (I think) with all of the best rate conversion settings. (But now I'm questioning some of the settings and may have made some faulty assumptions) Let me research this and I'll see if I can get a better process.

Really well done video. Action isn't exactly the easiest thing to try and pull off and you (and your team) created a well-shot, believable piece with some great cinematography. Very nice. And thanks too for going the extra mile and posting all of the info about your settings and workflow.

I was wondering why, when you wanted to create a final video in 24P, that you chose to shoot so much of the footage at 720 30P and 720 60P instead using the actual 24P mode at 1080? I get for the slo-mo part but didn't you force yourself into doing a lot of extra conversions to get 24P when the camera shoots 24P natively in the first place?

Thanks for the positive comments - just wait the scene goes nuts from here.

As for the why the different modes thing (and this is my bad for assuming people would know thism, so I'll ad it to the top of the thread) but it's been fairly well documented that while the GH1 has mightily conquered the skew issue (it's there but minimal) that DSLR cameras suffer from, it's introduced a new (probably worse issue) "Mud" or image clumping / macro blocking under extreme motion in the AVCHD modes. If I could shoot 108024 for everything I would be a happy camper. But theoritically it was impossible to do what I have done to the degree that I've done it - intense action shooting with the camera. In 1080 24 AVCHD the "mud issue is worst. In 720 60 AVCHD it's there but not as bad. In 720 30 MJPEG, it's not there. This is because the AVCHD modes are Long-GOP meaning the camera tries to only redraw the changing parts of the image. With too slow a bandrate it too much of the image changes between redraws and creates weird clumping. The Mjpeg is Intraframe - it takes 30 progressive indivudal pictures per secon redrawing the whole image every time. In AVCHD maybe the 60p has less mud because it's a smaller image to redraw with more samples? Dunno, but that's the way it works.
SO

• 1080 24 AVCHD - most resolution, best frame rate, most mud with too much movement.
• 720 60p AVCHD - less resolution, less mud, useable for most
• 720 30p AVCDH - less resolution, more compression artifacts, no mud

So the strategy for shooting intense action stuff suggests istelf somewhat (provided a good frame rate conversion method can be implemented) and goes as follows:

• For wow resolution shots, look at that landscape, look at that whiskers on the character actors chin, gee what quality without too much movement - wow 'em with 1080 24p.
• For a scene with average movement, crossing to the fence and back, you can use 720 60p AVCHD. You've already established the "wow what quality!" feel with a few 1080 shots and this mode will serve you well
• But fling the camera around like a madman, and you must be in 720 30 MJPEG. To my eye there's a bit of line sharpening in the MJPEG mode, but it's minimal, can be minimalized further by reducing sharpening in camera, and won't be noticed if you're flinging the camera around like a madman. But for beauty shots use 1080 24, for compromise between beauty and motion use 720 60, for best motion in extreme circumstances use 720 30.

Nice. That's really good to hear. I'm guessing ProRes422 would be the best codec to use?

I've been using CineForm. Smaller file size than Pro Res, 422 colorspace and less noisey. ProRes adds a bit of noise.

If you're on a Mac but the PC people are going to feel a little left out. DVCPro HD would be more platform universal.

From what I understand DVCpro HD has been implemented differently on Mac and PC and doesn't work well cross platform???

I thought you couldn't play DVCProHD files on a PC without some special software that would need to transcode to a different codec. When did this change? Admittedly I haven't used a PC in years. What about H.264 at best quality? Good enough?

Then I guess the DVCProHD codec should work then if you keep the same 1920x1080 resolution and pixel aspect ratio of the AVCHD files.

After discussing with Barry Green what I'm going to do is upload BlueRay quality H.264's. Everybody should be able to play these, and if BlueRay quality ain't enough then we're screwed

No such thing as to clean............ I would rather start out noiseless ( on the senor level ) then add through ISO or post to dirty it up. To have the choice is better then to not to have one; every and any time!

I tend to agree though I've always found the HVX noise to have a fairly organic pleasant feel and not distract from the image (within reason - never shot at gain +6 with no light, etc.) And yes even some film stock is so clean it has to have grain added in post for a filmmic look, lol.

Still confused why your footage ( large downloaded file ) looks more noisy, videoish than ( comminderspike' ) night footage shot at 400-800-1600 iso in 1080/24 mode.

Well maybe it has a lot to do with the rate conversion. He's shooting in 24p - perhaps I'm not getting a good simulation of 24p in my post process. Also he's picking beauty shots - and is he shooting with lenses besides the stock lens? I've done my best to make a pleasing image, but I'm shooting in a crappy apartment complex with tan walls and a blown out overcast sky trying to follow the action with a near infinite DOF and trying and get this whole thing shot in 5 hours. I would say the shots of the kids and people in front of their hotel is pretty filmmic. That's got SDOF and real 24p and looks pretty damn filmmic to me with the nice backlight and the nice way CMOS blows out etc. But I would say not cherry picking my shots, mixing frame rates (perhaps with poor rate conversion in post) and being near infinite DOF following action certainly sounds like a recipe for less strikingkly filmmic.

Blooms desert piece as well looks more filmic and that was in 1080 mode. Khoile GF shopping spree vid and hunters stills also look more filmic. I think your having some kind of pulldown artifacts too unrelated to the above obsevations...

Bloom's desert piece was shot with filters and lenses that cost 4 times the price of the entire camera EACH. Kholi was also using fast primes. I was using stock lens. I'd put the shots of the kids beyond the fence up against the other footage you just mentioned. My point is that those guys weren't running around with the stock lens covering stuff doc style.

Also we'll have to see about the post conversion. As I mentioned, looking into that more now ...

same here! also the 720p60 parts use 1/50 so its really 720p30, exactly like mjpeg parts, the stuttering seems to be related to 30p-24p conversion.

better use a HDV at 24p for these parts... more quality and the DOF is the same.

He clearly explained what happened with the 1/50. HDV? LOL.

Yeah that's gone over in the first post and on a title card in the video in big block letters.

BUT .... BUT .... BUT .... What if you are shooting 720 60p AVCHD witha a shutter of 50 (as I did) and this forces the camera into 30p (as the folks at NeoScene confirmed for me) .... might this not be a anti mud AVCHD hack? Perhaps? Less image than the 720 to deal with, you've haved the frame rate from 60 to 30 maybe thereby doubling the efficiency of the 17mbs bandwidth ... could this be? Could this be why I had almost no mud issues? Could this be why I did have mud issues when I went back and did my reshoot at 720 60p? I don't know. There could verywell be a flaw in this logic or ... perhaps it's so crazy it just might work! Barry Green and I are sharing a camera right now and it's his turn to get it back. If there's a mad genious that could confirm or debunk such a crazy theory it's him.

Thanks for your interest and support all. Will have more shortly hopefully. This is fun and exciting to be on the cutting edge, be able to share what I'm learning and learn from you guys in this uncharted territory from me. I'm 90% of the way convinced that the cost to bag to art to usibility ratio of this camera is incomparible. If we keep the back and forth SOLUTION oriented as all of you guys are I'm optimistic about bridging that last 10%.

:beer:

 

[Jack Daniel Stanley]

One more thing, the 30P MJPEG footage I've seen looks like it has some sharpening applied at default settings. I've seen this on some other 30P shots from the GH1. I'm curious to see MJPEG with sharpening dialed all the way back.

Isaac all of my 30 MJPEG has the sharpening dialed as low as possible, plus contrast as low as possible and shot in the "SMOOTH" preset for "low contrast and smooth slightly destaturated colors". I've added some sharpening in the color graded section, but the raw section at the end fits your bill.

 

[Isaac_Brody]

Gotcha, will rewatch that clip. Looked good to me, just noticed this on some other clips on vimeo.

 

[Jack Daniel Stanley]

I definitely still see some line sharpening around faces, but I see it in other modes too - maybe as a result of hacking the camera in to 30p in the 60p mode due to my 1/50th shutter????

But 30MPJEG's great for intense motion when you won't have time to notice the sharpening.

 

[Mike@AF]

I don't know about distracting - if we're seeing the same thing anyway - because I've watched it over and over again and I'm only 80% sure I even know what you guys are talking about. Barry Green says he's not seeing it in the 200mb file either.

But I kind of think I'm seeing it now and am researching the 30p tp 24p thing and will try a new method and detail the new and old methods shortly.


Although I didn't do it by dropping frames. Which is what it would be if you just dropped it into a 24p timeline. I did it by interpolating to 24p via compressor (I think) with all of the best rate conversion settings. (But now I'm questioning some of the settings and may have made some faulty assumptions) Let me research this and I'll see if I can get a better process.

What I'm noticing with the stuttering can be seen in a frame by frame look at the shot of the camera moving along the fence at about 00:00:18. When you go frame by frame the distance the camera moves jumps about every 4 frames. I'm guessing this is something to do with the 1/50 shutter. I guess it could be the frame rate change as well and a look at the raw footage would tell us if that's it or not.

Personally I prefer the clean (or "too clean" look), and by clean I don't mean a real deep or infinite DOF. Too me it's more realistic. When you look through your eyes, do you see grain or noise? I'd rather have the need to add noise or grain or softness in post if necessary. Some may think too clean is too video-ish, but you can have a very clean look that still appears film-like. It comes down to proper composition, lighting, movement, DOF, etc. and then a good post process. This is all just me and my opinion, so I'm actually happy with what you've done here and think you did a great job for only 5 hours of work and being that the camera is new to you.

I just ordered a GH1 for myself and am looking forward to it, and this action sequence you threw together so quickly really helped push my decision over the edge. So thanks!

 

[DrBlaz]

He clearly explained what happened with the 1/50. HDV? LOL.

he explained that 1/50 was a mistake, but not that using 1/50 gives 720p30 instead of 720p60


HDV? LOL ?

could you explain us why its better to use mjpeg at 30p than HDV at 24p for a 24p video?

 

[Mike@AF]

Also, there's two frames (00:00:27.36 and 00:00:27.40 in QT). These are consecutive frames and it just seems like his leg traveled far for one frame so it seems like a frame was dropped/skipped.

 

[DrBlaz]

BUT .... BUT .... BUT .... What if you are shooting 720 60p AVCHD witha a shutter of 50 (as I did) and this forces the camera into 30p (as the folks at NeoScene confirmed for me) .... might this not be a anti mud AVCHD hack? Perhaps? Less image than the 720 to deal with, you've haved the frame rate from 60 to 30 maybe thereby doubling the efficiency of the 17mbs bandwidth ... could this be? Could this be why I had almost no mud issues? Could this be why I did have mud issues when I went back and did my reshoot at 720 60p? I don't know. There could verywell be a flaw in this logic or ... perhaps it's so crazy it just might work! Barry Green and I are sharing a camera right now and it's his turn to get it back. If there's a mad genious that could confirm or debunk such a crazy theory it's him.

Jack, no way! as said before panasonic did a good job crippling the encoder, when using 1/50 and 1/30 shutter speeds in 720p60, the encoder uses only P frames to reduce quality (in 720p60 it uses an I frame every 30 "bad" P frames), at least it is so in all the 1/30 mts's I've downloaded

 

[Jack Daniel Stanley]

he explained that 1/50 was a mistake, but not that using 1/50 gives 720p30 instead of 720p60


HDV? LOL ?

could you explain us why its better to use mjpeg at 30p than HDV at 24p for a 24p video?

What I said was:

Anyway we shot 1/50th shutter and this forces the camera in to some kind of hack for the 60p. Either it forces the camera into a lower frame rate / and or it drops half the resolution.

I've amended it to:
"... Either it forces the camera into a lower frame rate (presumably 30fps) / and or it drops half the resolution."

But we don't know exactly how it's hacking to get the 1/50th shutter. Do we? The people at CineForm confirmed for me that my 1/50th shutter 60p clips were 30p in a 60p wrapper though.

Jack, no way! as said before panasonic did a good job crippling the encoder, when using 1/50 and 1/30 shutter speeds in 720p60, the encoder uses only P frames to reduce quality (in 720p60 it uses an I frame every 30 "bad" P frames), at least it is so in all the 1/30 mts's I've downloaded

OK. How are you arriving at this? Just clicking through? I'm not saying your wrong - really asking. I've said it forces a lower frame rate or drops resolution - but we really don't know yet. I DO know that I got less mud at 720 60p with 1/50 than I did at 1/60. I said thinking of this is a mud hack was a crazy idea, but I don't see how your point suggests it couldn't be a crazy idea that might work.

 

[John Caballero]

could you explain us why its better to use mjpeg at 30p than HDV at 24p for a 24p video?

HDV, thats a fading format. Got to look forward.

 

[DrBlaz]

What I said was:

But we don't know exactly how it's hacking to get the 1/50th shutter. Do we? The people at CineForm confirmed for me that my 1/50th shutter 60p clips were 30p in a 60p wrapper though.

well, I haven't the camera yet, so all the info I have is from downloaded mts's, when using 1/50 and 1/30 in 720p60, the GH1, like other video cams when using shutter speeds over framerate, records repeated frames, it is not 30p in a 60p wrapper, its just repeated frames. If the GH1 allowed 1/15 then you'll get groups of 4 repeated frames.

OK. How are you arriving at this? Just clicking through? I'm not saying your wrong - really asking. I've said it forces a lower frame rate or drops resolution - but we really don't know yet. I DO know that I got less mud at 720 60p with 1/50 than I did at 1/60. I said thinking of this is a mud hack was a crazy idea, but I don't see how your point suggests it couldn't be a crazy idea that might work.

If you examine the mts with any h264 analizer you can see the mts structure. The gh1 does NOT use b-frames (all $500 avchd cams including panasonic use b-frames), 17mbps is too low for 1080p24 and 720p60 without b-frames.

When recording 720p60, the gh1 uses a I frame every 30 P frames. If selecting 1/50 or 1/30 shutter then we get 30 repeated frames (or what is really interesting, 25p on the pal model) , if the encoder structure is the same, then we could get in theory 720p30 or p25 at 17mbps, this is, the same bitrate for half frames (like panasonic P&S photo cams), so the mud could dissapear.

But, it seems panasonic thought about this mud workaround , and it seems when recording at 1/50 and 1/30 in 720p, it changes the way the encoder works by eliminating all the I frames, this is, only an I frame at the beggining and then all P frames, also it seems to use a lower bitrate, so the mud is still there. This is my "conclusion" from the few mts I've downloaded, I think all were 720p60 at 1/30, if you upload a 1/50 original mts I can check it.

But anyway, the only mud workaround would be in the PAL model, cause to get mud free images at 30p, you better go with a 5dII.

I just requested in the other forum 720p PAL footage at 1/30 to check it , we'll see, but probably panasonic crippled it also.

 

[Mike@AF]

If you're on a Mac but the PC people are going to feel a little left out. DVCPro HD would be more platform universal.

After discussing with Barry Green what I'm going to do is upload BlueRay quality H.264's. Everybody should be able to play these, and if BlueRay quality ain't enough then we're screwed

Actually you can go with ProRes. Apple has the codec for windows here:
http://support.apple.com/downloads/Apple_ProRes_QuickTime_Decoder_1_0_for_Mac