Thread: Slow Motion with RED : Best practices

Reply to Thread
Page 1 of 7 12345 ... LastLast
Results 1 to 10 of 63
  1. #1 Slow Motion with RED : Best practices 
    Moderator David Battistella's Avatar
    Join Date
    Apr 2007
    Location
    Florence, Italy
    Posts
    9,999
    Hi, There are a lot of new people coming to the forums and I thought I would start this best practices thread as a resource to people shooting high speed footage with their RED cameras for the first time. Want to welcome those with anything to add to please contribute to this thread.

    A RED camera treats over cranked footage much like film. This is how it works:
    • The timebase is 23.98 (film is 24)
    • The number of frames you shoot in the time base is the over or under crank. Let's use 72fps as an example.
    The RED captures 72fps (in 23.98 timebase) instead of the regular 24 fps in (23.98 timebase) so that when you play back the footage in a 23.98 timeline, you see it plays the 72 frames per second in a 23.98 time base the way film does. For example, if you record for one second at 48 frames per second it takes two seconds to play the footage back. Here is a chart to help illustrate this:

    Slow Motion: time in seconds of footage recorded at different fps compared to acquired time in seconds of footage in slo mo.
    • 1 second @24 FPS= 1 second clip duration plays back for 1 seconds in a 23.98 timeline
    • 1 second @48 FPS= 2 second clip duration plays back for 2 seconds in a 23.98 timeline
    • 1 second @72 FPS= 3 second clip duration plays back for 3 seconds in a 23.98 timeline
    • 1 second @96 FPS= 4 second clip duration plays back for 4 seconds in a 23.98 timeline
    • 1 second @120FPS= 5 second clip duration plays back for 5 seconds in a 23.98 timeline
    It is evident that with RED or FILM, over cranking the camera is actually capturing more frames in one second. This is the exact principle the SUPER high speed cameras like Weiss Cam and Phantom use, as well. These cameras are capable of capturing up to 1000 or even 4000 frames in one second. That means (with enough light) they can "see" a bullet flying through the air.Film works in exactly the same way. It captures more information in the same time base. One second is always one second, you are just telling the camera to capture MORE FRAMES in that one second, so in film and RED you are simply capturing more information and you do not need to do anything other than drop the footage into a 23.98 time base.What happens in a 5D (or Panasonic HVX 200) is that they use the sensor's 60 frame clock speed to capture the extra frames. While the footage is being recorded, frames are marked for processing. Through a special process, the footage is extracted and the frames playback as slow motion in a 23.98 timebase. (99.9% of all video based editing is 23.98 and NOT 24fps).

    The way RED differs from film is this.The RED reduces the amount of sensor you capture in order to get higher speeds.

    Here you can see the frames per second increases as you window the sensor (capture less image area.













    LEAVE YOUR TIMEBASE ALONE JUST CHANGE THE FPS
    To many people think that they have to use a 59.94 timebase for slow motion. DON'T DO THIS. Another in-camera method is to capture at 59.94. You can to convert 720p60 footage into very smooth slow motion using the CONFORM option of Cinema Tools. My advice when capturing 59.94 is to capture progressive frames; this way the footage can be interpolated back to 23.98, a pulldown can be added and a slow motion effect can be created by telling the software that the footage is 23.98. When it sees 60 frames, it plays them back at 23.98 speed. Here again, the limitation is 60fps slow motion. If you wanted to make that 120fps slow motion, you would only be doubling frames instead of playing back 120 unique frames -- you would be playing back every frame twice to simulate a slow motion effect.There is a big difference between in-camera Film Style Slow Motion and this method. Remember that with film style slow motion, the camera is capturing every frame, whereas in post-produced slow motion, the frames are being interpolated.

    IMPORTANT FACTORS - CHECKLIST
    The challenges with high speed is three fold.
    1. You loose stops as you request higher speed
    2. The RED needs to shoot at higher compression rates to shoot high speed (it's a. Bandwidth issue)
    3. To get to the Highest speeds you shoot on a windowed sensor.
    4. Shutter Angle can change requiring Black Shading calibration for the shutter speeds you shoot at.
    Many image quality things can work against you.

    THE COMMON MISTAKES
    underexposed/high ISOhigh compressioncropped sensorNo new blackshading

    REMEMBER STOP LOSS WHEN OVERCRANKING
    24fps = 1/48 (benchmark light level)
    48fps = 1/96 (1 stop light loss)
    60fps = 1/12096fps = 1/192 (2 stops light loss)
    120fps = 1/240
    192fps = 1/384 (3 stops light loss)
    240fps = 1/480
    300fps = 1/600 (just under 4 stops light loss)
    768 fps = 1/1536 (5 stops of light loss)
    1,000fps = 1/20001,536fps = 1/3072 (6 stops of light loss)

    R3D SAMPLES from CLIPS

    Vimeo https://vimeo.com/163683335


    Last edited by David Battistella; 04-21-2016 at 11:40 AM.
    "Colors are the wounds of light."
    -William Blake
    Reply With Quote  
     

  2. #2  
    Moderator David Battistella's Avatar
    Join Date
    Apr 2007
    Location
    Florence, Italy
    Posts
    9,999
    One of the most important things to understand is that the DRAGON sensor is completely different from the MX sensor. It is a new imager and it needs to be tested as thought it is a whole new camera, because, in essence it is.

    The STH OLPF combination with the Dragon sensor in particular is geared toward skin tone/drama work and creates a base image which most resembles the Alexa (somewhat green tint). You are seeing the combination of a few things I described in an earlier post. STH, underexposure, windowed sensor, high frame rate and black shading calibration.

    With the STH it's a great plan to rate (depending on the scene) lower ISO and hit the sensor hard (with light). This scene you present on STH (with a lot of high shadow area and wide contrast) could have been rated at 250-400 ISO and this would ensure that the you would not be seeing RED DOT shadow noise in the image. You can really push the dragon sensor with a lot of light before clipping with the STH installed and I would say ESPECIALLY in scenes where you want to get shadow detail.

    I think the same shot on Standard would not be producing this effect. The shadow areas can get a bit mushy in lowlight.

    We are so used to protecting highlights with the MX sensor and digging stuff out of the shadows, that if we expose this way on the STH filter, you end up with the kind of artifacting you are seeing here. IN my first days with the DRAGON STH combo, I had some similar exposure stuf happening, but I noticed it less because I always rated MX at 320 and did so with the DRAGON STH combo. I noticed great improvements, but others (who tend to use a base ISO of 800 we getting caught with RED dot noise due to under exposure. IF you rate the STH at 250-320 open until the image clips and wind it bac a half stop, you are golden.

    The LLO and STANDARD OLPF's give you a much more similar image to the MX, with the benefits of the DRAGON sensor.

    For 253 FPS you had to be in 2K and, just as a reminder, you are using less that 1.4 of the sensor resolution to capture your images (that is the trick to RED slomotion) it is not scaled but windowed, by exposing only 1/4 of the sensor, you can UP the framerates.

    Honestly, 3K or 4K slow motion can yield much better results because you are not punching in so much.

    See the samples in the post above.

    David
    Last edited by David Battistella; 04-21-2016 at 07:28 AM.
    "Colors are the wounds of light."
    -William Blake
    Reply With Quote  
     

  3. #3  
    Moderator David Battistella's Avatar
    Join Date
    Apr 2007
    Location
    Florence, Italy
    Posts
    9,999
    Quote Originally Posted by C. Phillips View Post

    Now here is the kicker, I posted this earlier and some of you watched it. But what you may be shocked at if you are honest with yourself, the opening shot is actually a 3K 3072 x 1284, 201fps SCALED UP image! If I didn't say anything, some of you wouldn't have known IF you're being honest. Still looks great! Redcode is 7:1. I forgot it was that was the resolution until I actually dug up the specs.
    God bless!

    -C. Phillips
    One of the most important things to understand is that the DRAGON sensor is completely different from the MX sensor. It is a new imager and it needs to be tested as thought it is a whole new camera, because, in essence it is.

    The STH OLPF combination with the Dragon sensor in particular is geared toward skin tone/drama work and creates a base image which most resembles the Alexa (somewhat green tint). You are seeing the combination of a few things I described in an earlier post. STH, underexposure, windowed sensor, high frame rate and black shading calibration.

    With the STH it's a great plan to rate (depending on the scene) lower ISO and hit the sensor hard (with light). This scene you present on STH (with a lot of high shadow area and wide contrast) could have been rated at 250-400 ISO and this would ensure that the you would not be seeing RED DOT shadow noise in the image. You can really push the dragon sensor with a lot of light before clipping with the STH installed and I would say ESPECIALLY in scenes where you want to get shadow detail.

    I think the same shot on Standard would not be producing this effect. The shadow areas can get a bit mushy in lowlight.

    We are so used to protecting highlights with the MX sensor and digging stuff out of the shadows, that if we expose this way on the STH filter, you end up with the kind of artifacting you are seeing here. IN my first days with the DRAGON STH combo, I had some similar exposure stuf happening, but I noticed it less because I always rated MX at 320 and did so with the DRAGON STH combo. I noticed great improvements, but others (who tend to use a base ISO of 800 we getting caught with RED dot noise due to under exposure. IF you rate the STH at 250-320 open until the image clips and wind it bac a half stop, you are golden.

    The LLO and STANDARD OLPF's give you a much more similar image to the MX, with the benefits of the DRAGON sensor.

    For 253 FPS you had to be in 2K and, just as a reminder, you are using less that 1.4 of the sensor resolution to capture your images (that is the trick to RED slomotion) it is not scaled but windowed, by exposing only 1/4 of the sensor, you can UP the framerates.

    Honestly, 3K or 4K slow motion can yield much better results because you are not punching in so much.

    See the samples in the posted in the thread I mentioned

    David
    "Colors are the wounds of light."
    -William Blake
    Reply With Quote  
     

  4. #4  
    David,

    These are good clips. I will try to lower the exposure to 320 and test again. Do you have any 200fps clips? I still would like to see the results at that frame rate/compression. However, these clips are Tremblant of the quality I was getting out of my Epic-X camera which I was perfectly fine with. Don't know what is going on with that Scarlet-W.

    Sincerely,
    Reply With Quote  
     

  5. #5  
    Moderator David Battistella's Avatar
    Join Date
    Apr 2007
    Location
    Florence, Italy
    Posts
    9,999
    Yes.

    I think I have some stuff of honeybees. I'll post into here later, but after a lot of initial testing I tend to stay away from 2K altogether.


    David
    Last edited by David Battistella; 04-21-2016 at 08:13 AM.
    "Colors are the wounds of light."
    -William Blake
    Reply With Quote  
     

  6. #6  
    Senior Member
    Join Date
    Feb 2016
    Posts
    197
    C. Phillips,

    I know you've gotten alot of heat over the last few days, but I do think you should ultimately be commended for starting this thread. Alot of useful information and enlightening discussion (borderline bullying aside) came out of the post. There's been a few very well-articulated bites of advice and information that I know I can carry over into my next shoot on these cameras. I normally rent. But I decided to pull the trigger on a red this time around because aside from the alexa and amira, the epic has always been our biggest and most common rental cost, so thought I would finally just try owning one. And these types of discussions are very useful for dp's who don't shoot exclusively on Red. Like someone posted earlier, it isn't in fact common knowledge that although certain features are advertised in the specs, they're not necessarily ideal or encouraged to use. So cheers for posting, putting up with the grief, and being constructively critical. Would be a bad place if a major manufacturer released a camera that had problems and no one dared to speak up about it. Though I don't see this as a problem with the camera, rather just limitations on certain shooting scenarios. But all the same, it's always best to speak up. So thanks for that.

    Quote Originally Posted by C. Phillips View Post
    Brody,

    No problem fam! Glad it all makes sense now! That was the reason for me even starting this thread, to help people make informed decisions to let them know what they are getting into and help with their expectations. I had the Magnesium preordered at first but changed thinking the difference shouldn't be that bad. But, without seeing any examples out there, I made the wrong decision. People keep getting excited with the Raven "Carrion" video which is REALLY dope, but all of those shots are probably no higher than 48-60 fps which would maintain a pretty low compression at those frame rates, but we still didn't see what it would do beyond that.

    Now, I am wondering if the Magnesium would be better. That $20,000 bump is just really high. If you get your Magnesium, I would love to see a quick clip at these frame rates.

    Thank you for the feedback! Glad you feel better about your choice in purchase now!

    Sincerely,

    C. Phillips
    Reply With Quote  
     

  7. #7  
    Senior Member
    Join Date
    Feb 2016
    Posts
    197
    Thanks for taking the time to post David.

    This is true however, that Red isn't entirely identical to film in this respect. The sensor (or gate) windowing is something unique to Red's HFR acquisition. So for people coming from Sony, Canon, Panasonic or Arri, it's important to understand what the limitations of overcranking on Red cameras are, and why this leads to inherently more noise at lower resolutions than what the might otherwise be used to. What I haven't researched yet is, what is the advantage of windowing the sensor as opposed to scaling, like sony for example. Anyone with knowledge or links are very welcome to chime in here. Thanks.

    Quote Originally Posted by Tarquin Cardona View Post
    David,

    The statement that RedCamera treats over crank exactly like film is not precisely true. I only point this out because you made a highlight of it, and it may need to be clarified to avoid further misunderstanding. With film, when you shoot high speed, the same quality and resolution of film stock simply passes through the gate faster. So you are getting many more frames of high quality image in the same time. The slo mo from film is therefore very good and clean. The way Red achieves high speed - because it is computational - is by gradually cropping the film stock (if you will) which is part of why we have these grain discussions. In truth, in order to process the temporal data the acuity of the digital film stock keeps being diminished through increased compression up to the limits of the internal boards.

    I think most of what you map out here is super helpful. I just fear people are going to continue to misunderstand how Red does slow motion versus film.
    Reply With Quote  
     

  8. #8  
    Senior Member
    Join Date
    Jul 2010
    Location
    Stamford CT
    Posts
    3,080
    If anything you may have talked me out of doing a raven deposit... I wanted it for low res hfr shots without taking the financial hit of taking my rental units out of rotation. But I may just wait for Arri to release a new sensor now instead since soon after that my Amira will be retired and realistically I'm too busy anyway currently to shoot much personal stuff. Or I'll see how much that new phantom mid range costs... Or move up to scarlet w... If need be for anamorphic...

    ps what about hfr on scarlet w monochrome? How much noise and artifacts will be in that model? It would be cleaner I imagine?
    DIT, Assistant Editor, Phantom Tech, offering workflow services from acquisition to post

    Available with or without following gear:
    Alexa Classic
    2 Alexa Studios
    Alexa M package
    Arri Amira Premium/UHD
    Alexa Mini with raw, 4/3
    Helium (Monstro deposit)
    Epic W
    Oconnor 2060


    Red Rocket X with rMBP, FCPX, Avid MC8, and Resolve Lite 12 in the NYC/CT area (but willing to travel).
    Reply With Quote  
     

  9. #9  
    Senior Member Daniel Reed's Avatar
    Join Date
    Oct 2011
    Location
    SF Bay Area
    Posts
    790
    David, thanks for making this thread.
    Benchmark username dread64
    Third fastest Mac Pro on GeekBench: https://goo.gl/e3ZZWK
    #1 for 12-core Mac Pro single-core on GB: https://goo.gl/ONW3qt
    #13 on LuxMark Medium (Win/Mac/Linux) CPU + GPU: https://goo.gl/gMJRdR
    Reply With Quote  
     

  10. #10  
    Senior Member Scot Yount's Avatar
    Join Date
    Dec 2014
    Location
    Marblehead, Massachusetts
    Posts
    1,292
    Quote Originally Posted by Kris Tait View Post
    What I haven't researched yet is, what is the advantage of windowing the sensor as opposed to scaling, like sony for example. Anyone with knowledge or links are very welcome to chime in here. Thanks.
    If you think of the data being written to your media card in volume/time, it is easy to see why RED allows you to window the sensor. For each frame at full resolution, every pixel on the sensor fires so to speak, and produces data. At 24fps at full resolution, call it one full bucket of data. If I want more frames per second, and my bucket is full, then I may choose to window the sensor, firing fewer pixels per second. I can then increase the frame rate right up to the point where I am again getting a full bucket. Compression plays a large role here as well. If my Scarlet-W writes 140MB/s to my mag, then that is the limit....RED is allowing me to choose how to use that data rate. Scaling by contrast, would fire every pixel every time, and then have to be scaled, further burning bandwidth and processor.

    It is pretty simple really. Sony and Canon don't want to have to explain to you why they gave you the tools to make a bad picture, so they don't. RED let's you jump into the deep end expects you to find your own life preserver. That's what RC and no in-camera sharpening is all about really.....
    Last edited by Scot Yount; 04-21-2016 at 03:51 PM.
    Forged CF Helium 8K 00107
    Scarlet X 02249 (The Gateway Drug)
    Fujinon Cabrio 19-90 Version 2
    RED Pro Zooms and Primes
    Director/DP at https://www.butcherbox.com

    Living life by the ocean.
    Reply With Quote  
     

Posting Permissions
  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts