Jump to content

Photo

anyone try ralph's Sanity 5 hack?


  • Please log in to reply
4 replies to this topic

#1 gsxr310

gsxr310

    Newbie

  • Members
  • Pip
  • 6 posts

Posted 16 August 2012 - 08:47 PM

just wondering as i've only seen talks of driftwood and flow motion on this board. from the replies on personal view, there are alot of praises of ralph's work. i've seen some of his hacks on vimeo and they do look good in low light situations. anyone here using this hack?

#2 sanveer

sanveer

    Advanced Member

  • Members
  • PipPipPip
  • 312 posts
  • LocationTravelling

Posted 18 August 2012 - 02:29 PM

I tried it once. It was decent. Actually I tried quite a few hacks. But, unless u're using a Huge Screen (Maybe 50 Inches or more), or can zoom into shots, and, or, use the footage for Colouring and Grading, the hacks don't really help you notice too much. The real difference can only be noticed, for technical purposes, and on keener observations.
The hacks are good. Any which ways.

#3 Mirrorkisser

Mirrorkisser

    Advanced Member

  • Members
  • PipPipPip
  • 183 posts

Posted 22 September 2012 - 09:29 AM

Its a very nice hack with small bitrates. I agree out of the camera you dont see too many differences between hacked and unhacked. Maybe less macroblocking in flowing water or a more natural movement. For those reasons i use one of the many settings outthere though. Super high bitrate Intra settings only make sense under special conditions and with a lot of postproduction work in mind.

#4 gravitatemediagroup

gravitatemediagroup

    Advanced Member

  • Members
  • PipPipPip
  • 142 posts

Posted 22 October 2012 - 02:56 AM

stay away from this hack
stick with flowmotion

from the creator of flowmotion himself when I asked about "make flowmotion have sanity 5's low light ability"


Low-level noise is generated by the image sensor and effectively amplified by the ISO setting. While a patch doesn't have the ability to produce more or less noise in an image, it can have an effect on how noticeable it is. In extreme cases, (such as the unhacked firmware) inadequate bitrate can smear the details in a way that makes them look less noisy. On a more subtle level, the Quantizer Matrices used in a patch can suppress the fine-grained distinctions in a region of noise.

Both Sanity 5 and Flow Motion v2 use individually customized Quantizer Matrices designed to suppress highly detailed chroma noise. This helps smooth out low-level gradients, such as shadows and dark skies. The drawback is that the lack of chroma details can sometimes make dark colors look blotchy. To minimize this tendency in FM2, I customized the luma matrices to preserve the fine details that are lacking in the chroma matrices. This technique correlates well with the eyes' greater sensitivity to luma details than to chroma.

Judging from the two patches' different Quantizer Matrices, Sanity 5 will suppress ultra-fine details in both chroma and luma channels relatively moreso than Flow Motion v2. In cases of very rapid motion, I've found that suppressing fine luma details can produce noticeable macroblock artifacts which are partially smoothed over by encoder's Deblocking Filter. My hunch is the unhacked encoder was tuned to work this way to compensate for its inadequate bitrate.

With Flow Motion v2, I had enough bitrate to eliminate almost all traces of macroblock artifacts solely by preserving the ultra-fine luma details. Since I no longer needed to use the Deblocking Filter to conceal macroblock artifacts, I was able to use it to smooth low-detail gradients instead. This is one of the core techniques that makes FM2 look and perform the way it does.

From a broader perspective, what this analysis revealed was an inherent trade-off between noise suppression versus artifact suppression. Flow Motion takes a zero-tolerance approach to macroblock artifacts, and is less concerned with suppressing low-level noise. This likely reflects my own preference to shoot at the lowest practical ISO setting using very fast lenses, rather than underexpose or use higher ISO settings that incur more noise. With largely static scenes, however, macroblock artifacts are less problematic, and underexposed areas may benefit from some noise suppression of the ultra-fine details.


#5 gravitatemediagroup

gravitatemediagroup

    Advanced Member

  • Members
  • PipPipPip
  • 142 posts

Posted 22 October 2012 - 03:00 AM

I tried it once. It was decent. Actually I tried quite a few hacks. But, unless u're using a Huge Screen (Maybe 50 Inches or more), or can zoom into shots, and, or, use the footage for Colouring and Grading, the hacks don't really help you notice too much. The real difference can only be noticed, for technical purposes, and on keener observations.
The hacks are good. Any which ways.


have you used sedna?
it will show you a major difference in quality no matter what size screen ; )




0 user(s) are reading this topic

0 members, 0 guests, 0 anonymous users