@L1N3ARX I noticed the same problem in my shoot yesterday using flowmotion, so your not the only one that had this issue

Ah, ok. Thank you.

@LPowell

On the first page you wrote:

"...I've tested and confirmed these features to work reliably at 60Mbps in 1080i FH and 720p H video modes. The 60Mbps FH and H video modes also support 4GB file-spanning on 30MB/sec Class 10 SD cards, and produce excellent motion picture quality that is just a notch more compressed than Flow Motion's 100Mbps FSH and SH video modes."

For some TV productions I need to use the 1080i50 mode, which looks pretty good with FM2.02. I'm using a Panasonic lens and also much of these automatic camera functions. As you stated above, to be safe it's better to use the FH-setting than FSH-setting. I'm courious, what's the exact difference between those settings, is it only the bitrate or are there used longer GOPs or something like that?

Would you say with a fast SD-card (95MB/s) it is also safe to use FSH settings (incl. auto-functions, spanning etc.) or better stay at FH in that case?

And - last but not least - thanks for your work!

@LIN3ARX

This issue is much more pronounced once it's rendered to H264...

I didn't spot the issue in your YouTube video, all I could see was YouTube macroblocking in the shadows. Could you make the original FM2 footage downloadable on Vimeo? Generally speaking, once you render original footage to another format, the transcoding makes it difficult to identify how the frames were originally encoded.

Hi everybody, i am just a newbie , i upload the Flow Motion patch to my GH2 and i am not able to select the "record mode" when i am in manual movie mode ? ( like 720/60p ) Any suggestions ? Thank you in advance

@Butt love it !

@Vitaliy @LPowell @driftwood @cbrandin Hello Gentlemen, I must say that the strange reactions to criticism has begun to tear apart this adventure. Here in the great US we say opinions are like assholes, everyone has one. I think I have been here for maybe 2 years now and have continually been amazed at the camaraderie that exists here. Now it is competitiveness, the minions are gathering, pushing the individual teams to side. This is crap, are we forgetting what brought us here in the first place. It still blows my mind when it look back at ALL the patches and laugh at the beginning. I was knocked off my chair when I was testing Driftwoods 244, wow. When Vitaliy, would tell me wrong place, or read the wiki, it was good and It can be again. LLPowel ready for confrontation. Come on. Most of the posting here, no matter what patch is being used is more for what, Hey remember the fundimentals, Light, Composition, Lens , not this, "can you tell me what is the best patch for my job that I am shooting tomorrow"? Figure it out for yourselves, Test. These people have moved all of us forward,(@Vitaliy @LPowell @driftwood @cbrandin) made us wonder again. Opened our minds to the possibilities with this little cheap ass camera. I do not know where most of you came from, but I know what it is like to carry a full dress Panvision camera up a hill side, its is not fun. So please respect one another. Sorry Vitaliy, may be this is not the right place, it was the only one that might be read. Thank Guys, your great.

@flaschus

Thanks for your honest sarcasm. As for GOP-1 versus GOP-3 playback, the motion quality is crucially dependent on the playback device. In my own tests of Flow Motion v2, I have been consistently unable to spot any perceptible differences between its I-frames and its B-frames. The screen shot I uploaded at the top of this page is actually one of FM2's B-frames, compared to an I-frame from AN Boom Flat 4s.

One reason I haven't developed a GOP-1 all-Intra version of Flow Motion v2 is because it's impractical. The peak bitrate of that test video is around 100Mbps, and the average size of its B-frames is about 300Mb, plus 800Mb for the I-frames. To maintain that level of image quality, the GH2 would have to operate reliably at a peak bitrate of over 170Mbps, well beyond the maximum bitrate I've been able to achieve.

What's worse is that setting 1080p24 mode to GOP-1 forces you to run 1080p25 HBR mode at GOP-1 as well. With Flow Motion v2, I was unable to achieve reliable peak bitrates higher than 100Mbps in HBR mode, and that was with a GOP-3 setting. Changing that to GOP-1 would have significantly degraded the HBR image quality I was able to obtain at GOP-3.

@LPowell,,, I was being sarcastic in order to develop rapport. professor chaos.

@LPowell... I hope someday I can meet u so I can crack this stubborn streak u have...

bottom line, when i have a highlight on a face; or hot spot in the background, gop1 FEELS far smoother than gop3 in playback... ya? not a screenshot, but how it FEELS wheen played back on a screen. nomatter what.... everytime i have compared I can see the compressed B or P frames....

and the 1gop grades better.... like i have said to you before, I want a flomotion 1gop patch, for an HONEST COMPARISON

... ya dig? wheres the flomotion-INTRA????

@lpowell i like your comparisons. But please dont make them feel like an act of aggression.

I am also very curious how valkyrie 444 does compared to flowmo. Valkyrie is really great, just the file size rapes my hdds.

@valpopando Thanks for uploading your footage, it looks beautiful. If presenting independent, objective test results is considered an Act of War by Driftwood and other enthusiasts, then it's a "fight" I'm prepared to engage. I fully understand that the Apocalypse Now Boom Flat 4s settings are experimental, and so I chose a relatively undemanding scene to test it, rather than one designed to break it. I limited my evaluation to technical issues where its performance appears to fall short of its intended quality level. Hopefully, these test results will provide insight on how the patch can be improved to maintain a more consistent range of image quality.

@Mirrorkisser

[Apocalypse Now Boom Flat 4s] tries to be extremely sharp but then runs out of bandwith etc so in some parts of the frame the image quality really drops. It has to be said though that Nick always said its an experimental setting and it was not recommended for wide or super detailed shots.

That's right, I used Elecard Streameye to examine the macroblocks in an AN Boom F4s frame, and found that it was able to maintain its initial QP of 18 only in the top row of macroblocks. In the rest of the frame, QP ranged as high as 34, which indicates very coarse quantization. As Chris Brandin explained in earlier posts, the encoder will attempt to maintain QP consistent within a range of 5 around the initial QP level. In all cases where I've seen QP break out of this range, it was because the encoder was being forced to operate at an excessively high level of image detail that required more bitrate than it could sustain.

While it's true that AN Boom F4s was released as an experimental patch, it has generated a great deal of interest as Driftwood's latest offering. I was interested in it primarily as an example of using maximum bitrate to record as much image data as possible. My initial tests showed that it would fail consistently on highly detailed subjects in 80% Slo-mo mode, so I chose a simpler scene for this 1080p24 comparison test.

As can be seen in the Stream Parser report above, this is a scene that Flow Motion v2.02 encodes at its finest image quality, maintaining consistent QP levels within the encoder's standard 5-level range. That makes this FM2 example a good reference point for this test, since it's performing exactly as it was designed and intended to work. This is a relatively simple scene that I would expect any well-designed patch to handle without straining its resources.

Frankly, I was surprised to find that AN Boom F4s had so much difficulty encoding this scene at its intended quality level of 18. What that indicates is that its Flat 4s quantizer matrix is demanding far too much bitrate to operate consistently at a QP of 18. Since 150Mbps is close to the peak bitrate the GH2 can reliably sustain, my conclusion is that either the target QP of the patch needs to be raised significantly higher (i.e. coarser) than 18, or the Flat 4s matrix needs to be replaced with something more realistic.