HDMI Capture Problem SOLVED - AviSynth RULES!
- Thread starter Ralph B
- Start date
On Barry's clip they both leave duplicates if you use default settings. I more familiar with TDecimate, but I'll see if I can find working settings for FDecimate() and report back
Not sure what the celebration of "uncompressed" 4:2:0 footage is. In the other thread, some smarter than me had already concluded that the chroma smearing was probably due to the HDMI signal being from an inferior colorspace, trying to upsample into 4:2:2. It feels like we're celebrating an offroad truck that is doomed to spend its life on 12" rims. Seriously, what's the point?
That's why I stopped pursuing any solution on the other thread. It's 4:2:0 for crying out loud.
That's why I stopped pursuing any solution on the other thread. It's 4:2:0 for crying out loud.
Cosimo Bullo
Veteran
The EX1/3 (and coming sony f3) are 4:2:0 and we feel the codec in MUCH stronger than the avchd we're getting from the GH2 (haven't played with the AF100 yet). So we say, bring it on Ralph! And thank you!
Well if I might try to wipe up the urine you just left on everyones parade... I think that there's certain very real points to having this ability. Even if they're not as useful has having 4:2:2, it doesn't make them useless. First off, any quality gain, no matter how marginal, is going to be appreciated by a good handful. Having played with some previously posted HDMI footage myself, you do get a finer degree of detail and I might point out that a lot of the footage that's been posted so far aren't exactly codec killers.
There are likely many situations where even if the GH2's improved compression methods don't completely die and mudout, that the strain of high detail and motion can make the difference between the AVCHD and the uncompressed capture more pronounced. Furthermore, although this might be a smaller demographic, but for those doing a lot of controlled in-house work with the camera, being able to record straight to hard drive rather then dealing with SDHC card swapping, might be a plus. Think of screening work, even if the signal is 4:2:0, getting a minor jump in quality sent straight into your workflow without having to wait to transfer off of SDHC or perhaps bypassing the use of multiple card if it's long form... definitely not a negative.
vcassel
Veteran
You might want to come down off the soapbox and check the first post again.
Cosimo Bullo
Veteran
!!! Funny!
Hey PDR, I had a chance to apply your suggestions to some of my footage and I confirm that tdecimate (mode=1) is superior to fdecimate(). Using avisource and AssumeFPS(24000,1001) are also sound suggestions. I'm going to add them into the script in the first post. I would like to make that script as solid as it can be, so people can find it easily and not have to wade through a zillion posts.
I also tried your new settings for TFM, but noticed no difference. Could you explain your thinking behind them?
Oh, BTW, it's important to put converttoyv12() before TFM(). TFM requires YV12 or YUY2, and it's possible someone might record with a codec that is neither of these, in which case an error message will pop up. (I know, because it happened to me!)
Last edited:
Why so negative Man? I say never give up and get out of your sand box and play along. We owe people like Ralph a whole lot of encouragement and feedback to help make this and other functions work to our favour. Way to go Ralph!
nyvz
Veteran
While the difference in texture and detail is quite impressive, why show 60i comparisons (the AVCHD must be 17Mbps, right?)? Isn't your fix particularly for 23.98fps? I think it would be more meaningful to see the results from footage shot 23.98fps 24Mbps AVCHD compared to the same captured 60i uncompressed converted to 23.98fps progressive using your method. I'd be interested to see how much more detail is in 24Mbps footage from the GH2, since 60i 17Mbps would only give about 0.56Mb/frame versus 23.98p 24Mbps which is about 1Mb/frame, giving it about 75% more information per frame.
How easy is this conversion? I assume it would be as simple as running a batch conversion on all your footage? Have you tested that it maintains sync with linked audio if you record uncompressed from HDMI with an audio input? I imagine anyone using this method might be interested in how easily they can just run audio input to their external HDMI recorder and have it work properly throughout the whole workflow.
Last edited:
diegocervo
Well-known member
Ralph,
I just saw the images you uploaded in the first post and the result is amazing. Have you guys noticed how many more details are in shadows?
I'm so looking forward to put my hands on the atomos ninja!!
I just saw the images you uploaded in the first post and the result is amazing. Have you guys noticed how many more details are in shadows?
I'm so looking forward to put my hands on the atomos ninja!!
Danielvilliers
Veteran
I hear it a lot less but don't we get a lot of benefit from the uncompressed hdmi with the total dynamic range of the camera. If we can underexpose a bit more to preserve highlight and lift the shadows in post we get better dynamic range. The Avchd codec just kills the detail in the shadows as with the OP example so we can' t use this technique, but the hdmi output seems to hold much more detail in the shadows. It would be nice if during the test that the dynamic range is also taken into account. Not just sharpness or 442 because already the GH2 seems to give some good dynamic range, if we could extract some more that would put it well above the other DSLR not just in detail etc.
A very good test of this technique (I am not a big fan of HDR type of image but it tells you how much you can get back from the shadows) and benefit with the Sony F3, http://www.youtube.com/watch?v=l5la6uOumhM&feature=player_embedded I am not saying that the GH2 can do that but it would be nice to see how much can be extracted from a clean signal.
A very good test of this technique (I am not a big fan of HDR type of image but it tells you how much you can get back from the shadows) and benefit with the Sony F3, http://www.youtube.com/watch?v=l5la6uOumhM&feature=player_embedded I am not saying that the GH2 can do that but it would be nice to see how much can be extracted from a clean signal.
For the newcomers, I don't see this thread as new, but as an extension to the other thread I pointed to earlier. Ralph was there, towards the end, too.
And it was our discoveries over there that led Ralph to begin searching for a solution, which he has posted here.
So, that's not urine. It's sweat.
And it's not "everyone's parade." I was in that parade too.
Goading and prodding and downloading every piece of footage I could find off the GH2 when it first started to appear. I ask you, who was the first one to determine the cadence to the pulldown problems we were finding? Who determined that it was predictable, but within a range, and then correlated that to a buried comment in Philip Bloom's GH2 article, drawing attention to the idea that Panasonic engineers already knew there would be sync issues between the chip and the HDMI feed?
Geesh . . . I think I have earned the right to grumble a little . . . we were hoping for 422 uncompressed and didn't get it. . . what are your contributions to all this - besides flag boy?
What really cracks me up is that I had wondered if the HDMI feed was really just the AVCHD feed, months ago. Of course, that's worded wrong, so I was soundly spanked when I questioned it on these forums. Something smelled fishy, though, and I wasn't buying the whole sunshine and daisies until I could actually see it.
But the spirit of what I was afraid of was: are we getting an HDMI feed from the signal just before hitting compression on the camera, making it an inferior 420 color feed? And if so, why spend the thousands to get better compression, when our color resolution is being cut in half regardless? Even if you tether it to a computer, this is such a limitation I can't understand all the celebration.
But that's me. Heh, heh.
This isn't a very friendly place when you're not on a band wagon, is it? Heh.
Did you try it on Barry's clip? It makes a difference - the default settings will leave dupes.
The reason is his is a challenging one - it's basically a very close to worst case scenario. There is field switching which causes the TFM algorithm to associate the wrong field using the default mode=1.
The other clips I downloaded from the other thread didn't have runs like that, and the default settings were good enough
Good point , but I don't see the rationale in recording YV24 or RGB if you're going to end up converting to YV12 anyways (to get rid of the chroma sampling issue)? You're only wasting bandwidth and storage space. What happens if you set your capture card to 4:2:0 directly ? Are the artifacts still present?
Last edited:
Heh, easy there. I said everyone's parade not everyone elses... so that includes you.
I've been watching those same couple threads from about the get go as well, just didn't have anything productive to contribute.. although maybe I'll be able to do some fiddling now that we're in the AVISynth stage. You certainly have the right to grumble that 4:2:2 was a bust... I just don't get seemingly decrying all this effort as completely fruitless... that was all I was getting at, so I apologize if my statement inferred more
Also, thousands need not be necessarily spent if you don't need too portable of an option. Some people already own HDMI capture cards and computers that can be used for indoor work at no additional cost.
In case anyone doubts the value of shooting uncompressed HDMI at 4.2.0 vs AVCHD, take a look at these pictures. They were shot at 1080 24P, iso400, f16, 1/200 shutter, 14-140 lens @ 50mm.
I was not expecting the difference to be so dramatic. Personally, I thought AVCHD looked great and I could happily live with it. But, after seeing this, I'm going to have to rethink everything.
AVCHD recorded in GH2
HDMI captured with UT lossless codec at 4.2.0
This scene was shot through a dirty window. My capture card is in a desktop computer, so my range of mobility is somewhat limited. But perhaps it works out for the best - you can clearly see the difference in the darker areas and how AVCHD smears things up.
I was not expecting the difference to be so dramatic. Personally, I thought AVCHD looked great and I could happily live with it. But, after seeing this, I'm going to have to rethink everything.
AVCHD recorded in GH2
HDMI captured with UT lossless codec at 4.2.0
This scene was shot through a dirty window. My capture card is in a desktop computer, so my range of mobility is somewhat limited. But perhaps it works out for the best - you can clearly see the difference in the darker areas and how AVCHD smears things up.
Last edited:
Cosimo Bullo
Veteran
Amazing example, Ralph. Are you planning a white paper on this for us slower on the tech front?