A while back, I posted about my Live Streaming Studio V3.1 setup, because many people wanted to know what gear I’m using and how I get the “cinematic” look on a live Zoom call. To achieve that look, one of the things I had to learn from scratch was how to color grade.
BTW, do you need help with creating a great custom “look” for your film or video production, your camera, or your podcasting or stream? Give me a ping, and let’s talk. I sure as hell wasted an idiotic amount of time and money making all kinds of mistakes in the beginning, so I’m happy to help you avoid that.
Here, I’m sharing a bit about my further digging myself into a hole adventures into color grading with Blackmagic Design’s DaVinci Resolve Studio (free download of the non-studio version). It’s an incredible piece of software. (If you’re thinking about ditching Adobe Premiere – just do it! Go for it. I’ve never regretted it for a second).
This is not a primer on color grading. It’s just me dumping writing up and sharing what I’ve learned that works best for me so far. If you too wish to start (or continue) on a color grading learning journey with DaVinci Resolve, Cullen Kelly’s YouTube channel is probably the best place for that.
The following assumes you’re already familiar some of the concepts of color grading – or at least have a faint interest in how to create a cinematic image with digital tools. (If not, this post will bore the living daylights out of you is not for you.)
What started as a necessity during the lockdown era (aka building a professional-looking online tele-presence) turned into a rediscovery of my passion for the cinematic image. (Fun fact: Many might not know that I actually started out studying to become a film director – but I dropped out after only two years as university and studying film wasn’t really my thing. And then the commercial Internet happened.)
And as a person most likely somewhere on the spectrum, of course I can’t, I won’t, and I don’t stop digging until I’m getting somewhere interesting, somewhere where I can feel a sense of mastery and understanding of the full stack (lighting, lenses, camera, cinematography, sound design, microphones, color grading, post production, what not), aka being able to make predictable outcomes – and making those outcomes look cinematic and pleasing (to me). It’s become sort of a new obsession hobby of mine (in addition to helping other startup founders go farther faster, of course).
Still digging.. .
Read on below for this long non-startup (but still full of tech geekery) post.
The quick & dirty setup for the above shoot:
- Camera: A tiny (300g, w 8,23cm X d 7cm X h 6.6cm), old (launched in 2012!), and cheap (I paid less than EUR 600,- for it on ebay used, including an 8sinn cage, handle, and octopus expansion cable), digital super 16mm MFT dual gain sensor Blackmagic Design Micro Cinema Camera (MCC), ISO 800 (native), 5.600K, shutter at 180 degrees and 24 fps – obviously, exposed to the right (ETTR)
- Lens: A tiny (this being the largest in the series, but still tiny compared to e.g an EF lens) cheap (EUR 81,- on eBay, almost mint) vintage Pentax A110 (s16mm system) 70mm f2.8 fixed aperture (in this-lens-system-has-no-internal-iris! sense) on an MFT adapter
- Filters, etc: Lens is kitted with a 49mm metal lens hood that sports a 72mm ICE “IR/UV” filter (dirt cheap for the quality – and the MCC needs an IR filter if you’re shooting with any sunlight if you don’t want pink and purple blacks), a Lee Filters 2 stops IRND Proglass 100mm x 100mm filter (best ND filter I’ve ever used, IR filtering also helps immensely) on a cheap-ass Zomei 100m filter holder to shoot into the sun (classic backlit since I don’t own lights powerful enough to fight the sun) coming in at far side of my face (actually it was overcast and raining). No variable ND used as every single “variable” ND I’ve tested so far produces incredibly crappy color results.
- Lights: Key, Godox UL150 (silent, great value for money) with an Aputure Lantern modifier. Fill, Godox SL60 (not entirely silent, old-ish but OK – one of the best budget options for the color-accuracy at the time it came out) with an Aputure Light Dome Mini II softbox & honeycomb / grid modifier. (Thanks to Rob Ellis for being a fantastic source of inspiration, teaching how to make the most out of simple lighting setups and how to manipulate color temperatures to achieve the “cinematic” look even on a next to no budget!).
- Image Acquisition: Blackmagic Design Film Generation 1 DNG RAW (not to be confused with BRAW).
Below is a teaser reveal of my “The Creator” franken-rig, super 16mm ghetto style, that the above clip what shot with (the shot of the rig was done in the studio with the BMPCC4K).
Yes, of course I also couldn’t help myself from digging myself into another hole and obsessively over-engineering building my own camera rig to feed my compulsions fit my needs…
This franken-rig is shoulder and tripod mountable. On the shoulder it helps with stabilizing an otherwise jittery setup, and on the tripod, I can also remote control the camera with tracking (iOS app), and by joystick or 6-axis’ing with the remote (MasterEye) – and it features redundant hot-swappable power through v-mount batteries & d-tap.
The rig is so obnoxiously heavy that it has given me months of seriously unpleasant pinched nerves in the neck and god awful shoulder pains. Back to the drawing board. I’m now thinking about adding a Proaim Flycam steadicam vest. To a shoulder rig… Yes, I’m a lost cause. No, I don’t want to hear a word about the sunken cost fallacy at this point.
All of which amounts to an incredibly stupid amount of rigging for an old hd-only, tiny 300g camera.
But I am in love with the images that come out of that tiny old thing – when you get rid of the micro-jitters and grade it properly.
Let me know if I should do a video breakdown on the complete rig build & geek out further.
The rig jiggles a bit when rotated on the tripod in the video above, but I have later fixed this by adding a longer Manfrotto base-plate that helps tie more of the rig together, making it much more rigid. See the Kit.co page for details.
Since the last post, I’ve changed my Gamma output from 2.4 to 2.2 (because all I so far deliver for is online consumption and 2.4 is the old “TV” standard, 2.2 more in line with modern day phones, tablets, and computer monitors).
I’m now also using a “Video Monitor Lookup Table” by Cullen Kelly called “macOS Viewing Transform v1.3“, insuring that what I’m watching when grading is indeed as good as identical (good enough for non-pros like me, and still good enough for someone like me who has been working with pixels for +40 years and can spot by eye if one pixel differs 1 in value in any of the RGB values to the neighbours) to what gets delivered (YMMV if you don’t have a P3 Apple display – Mine is a P3 calibrated Dell 5K which uses the same LG panel as in the iMac 5K – afaik).
I also use an old Samsung SyncMaster calibrated to rec709 / Gamma 2.2 as the “CleanFeed” to compare to what I’m seeing in the main view in DaVinci Resolve.
Be aware that new color and gamma matching headaches awaits you when uploading your videos to the different content platforms like YouTube, Vimeo, Meta, TikTok, et al as they all have their own little wicked ways of re-interpreting (basically just ignoring) your color space / gamma metadata when re-encoding – and not all of them are publicly sharing what they expect to go in and what comes out of their sausage factories, making getting a predictable end result a trial and error riddle. In my brute-forcing experience with my platform of choice Vimeo (yes, I tried encoding and uploading using most non-obscure color spaces and gamma metadata settings and combinations thereof, and yes that took an idiotic amount of time to do, and yes the results were not exciting), exporting to rec709/rec709A is the most accurate fit to what I see when grading (YMMV – I do content for digital displays only, so far).
I’m also using an iPad Pro with the “DaVinci Monitor” app when grading. Just make sure the iPad is on the same WiFi as your Mac running DaVinci Resolve Studio – a weird & annoying limitation. (The Mac I’m grading on is usually only connected to the Internet & NAS/RAID via LAN, WiFi completely turned off, so I had a minute or two of frustration before finding out.)
<RANT>Once you get it up and running, don’t get me started on the incredibly stupid hassle of having to copy and paste the session access token string thing between devices when using the remote monitor – per f*ing session! This should be as easy as a click of the mouse, tap of the finger! I mean it’s all on the same network! I’m an adult, I can handle the security issues. Just give me the option to always allow when on the same WiFi network. If it’s good enough for Apple AirPlay, it’s good enough for me – and should be for you too, Blackmagic Design. </RANT>
Primary & Secondary Grading, My Clip-Level Nodes
Here’s my latest default clip-level node tree for primaries and secondaries – it works very well for me:
This node tree is almost copied almost verbatim from Cullen Kelly – and that’s because it’s an AWESOME framework that works very intuitively for me (too) – and disciplines me to keep things really simple. Top row = primary grade, second row = secondary grade (mostly masks of one sort of the other).
NR = Noise Reduction (this is the most CPU intensive node, placed at the start because of how proxies are generated and cached). RAT = Ratio or Contrast Curve. BAL = (Color) Balance. SAT HSV = Saturation in Hue Saturation Value mode (the node set to operate in a HSV color space and with only Channel 2 – the S for “Saturation” activated). Sharpen = Radius to taste (this node is placed after the primary and secondary grades to avoid potential artifacting that could happen you would put this node inside the primary or inside the secondary).
Lately, I’ve been using these Curve LUTs (esp. the “Film 2” LUT) to get the RAT node (ratio) 90% “right” (to my tastes) out of the box, adjusting the rest to taste depending on the clip. They’re made for DWG / Intermediate and don’t seem to break anything. Don’t forget to set the right pivot point for your color space in your ratio node (e.g. DWG/Intermediate = 0.336) if you want to push and pull it manually.
Not shown in my default node tree above: Sometimes I add the Sat Shaper+ or the Saturator (I like Saturator’s “preserve highlights” option) DCTL instead or in addition to the “SAT HSV” node I’m not completely satisfied with the saturation and sometimes also to modify the color density (yes, I’m lazy) and their “vibrancy” setting has sometimes helped me get more pleasing color separation or spread with one simple slider.
Sometimes I also use the TETRA+ DCTL if there are clips with some gnarly color issues that I’m just too incompetent to adjust otherwise.
I find myself more in the HDR wheels when adjusting exposure in the EXP node these days. I don’t know if that’s considered Kosher by the “pros” or not, but using the HDR controls for exposure feel much more intuitive for me – so I don’t really care.
My LOOK Node Tree, Timeline-Level Nodes
And this is my latest default timeline-level node tree for the “LOOK”:
BTW, you always want to be grading “underneath” your LOOK, aka always have your look nodes on the timeline level active (on) when you start grading your primaries on the clip level.
BTW, I don’t have internal grain activated in the Halation DCTL nor do I use the DaVinci Film Grain plugin much, as I mostly grade my own footage shot with the MCC that usually create all the graininess I need all by itself.
The idea behind the CST IN/OUT sandwiches is to be able to mix in creative (negative?) LUTs and Film Print Emulation (FPE) LUTs that were not made for the DaVinci Wide Gamut / Intermediate color space that I work in (the node directly in front of the sandwiches does take LUTs for DWG / Int should I happen to use one). In addition to my go-to look tool (here applied in the four nodes on the third line of nodes after DENSITY) which is Cullen Kelly’s beautiful Voyager Pro v2 “taste” LUT pack – worth every single penny!, I’m often adding in more creative or “negative” LUTs made for other color spaces, often – to the mix (literally, using the Key Input Gain of the node to limit its amount of influence) when I feel like. Often I use additional look luts from Arri (who doesn’t love themselves some Arri color science?). And my goto Film Print Emulation (FPE) is a Fuji 3510 by Cullen Kelly (free download available both for DWG & ACES, fantastic quality), and also sometimes I go crazy with Sony’s Technicolor Collection. For color density I’m using DRT&T’s Film_Density_OFX and Density+ alternately, as I’m still undecided which one I actually prefer. Dehancer is another fantastic plugin for creating the photochemical film look, but I keep it deactivated in the look node tree as I find myself wasting too much time trying to brute-force a look with it, as I’m still not very good at creating predictable results with it, haven’t really invested the time to learn how to use it properly yet.
Sidenote: Is there a DCTL / OFX plugin that ONLY does the FPE “analogue range limiter” part of Dehancer? That would make me happy. At least for a couple of minutes. Bueller… Bueller… Anyone?
Also deactivated by default is the Cullen Kelly YouTube export LUT. I normally use Vimeo for distribution, and like I mention above, I’ve found rec709 / rec709-A provides the best results when publishing on on Vimeo, aka looks most true to what I saw when grading after Vimeo has chewed on my upload and spat out their version.
There’s also a lazy “Global” node to abuse for anything I need to add as the last step for all clips, e.g. cool or warm it up a bit, take exposure up or down to taste, etc. – a handy node for quick and dirty experimenting with new ideas after I feel satisfied with the general look without touching the main nodes.
My approach for getting the look and feel I want is “less is better”, but anything goes (fuck around & find out!) – as long as I like it and it doesn’t break things (e.g. unpleasant skintones, artifacting, banding, etc), it’s a keeper.
My timeline look node tree also includes the MONONODES Balance and Clip utility DCTLs (so worth it!) and additionally I’ve added the native False Color plugin, as five “utility” nodes: False Color, Balance, White Clip, Black Clip, Sat Clip – and just by turning them on and off I can check the exposure and ratios, skin balance, and unwanted clipping across all shots (clips) really fast (turn the utility node on, select “refresh all thumbnails”, go to the Lightbox – and BOOM! – you’re checking all clips in seconds!).
UPDATE 1: Cullen Kelly just launched the fantastic “Contour” film look builder plugin that is now at the top of my wishlist. That is to say, this is a pro level plugin with a (fair) price tag that I’ll only allow myself to buy if and when someone will actually pay me for grading, lookdev, shooting, anything.
UPDATE 2: I’ve also added the new native DaVinci Resolve “Film Look Creator” as a default node (after Grain) with the “clean slate” setting to the LOOK timeline node tree.
UPDATE 3: I’ve also added a compound node with MTF emulation to the default LOOK tree. Because.
UPDATE 4: I’ve also added the 2499 DRT DCTLs to the default LOOK node tree, OFF by default, to provide an alternate DRT route and grade/look, as I experiment with it and evaluate it. Felt cute, might delete later.
Below is another grade I made after discovering the interesting 2499 DRT dctls from Juan Pablo Zambrano. This was my first try screwing around with them (looking forward to be playing around with the 2499 DRT tools some more):
Some more examples
Below you’ll find some more color grading examples where I’m going for the “super 16mm film” aesthetic, not the modern slick “shot-with-something-in-the-Sony-FX-or-Alpha-camera-family” look – because that’s my kind of kink.
Maybe I’ll share some of my more “modern” and “corporate-friendly” color grades shot on the BMPCC4K camera with a Sigma 18-35mm f1.8 Art DC HSM lens in a future post – for now, you can infer from my previous post on what that looks like together with this screenshot from my live streaming studio:
This is what the signal looks like before applying my custom studio grade and look:
Above, this is how it actually looks before you start color grading, color management bypassed in DaVinci Resolve. (BMD Micro Cinema Camera, DNG (RAW) Film G1, Pentax A110 70mm f2.8, ND6, IR Cut)
Color management on. Notice the slight bias to green I suspect snuck in with the K&F Concept ND filter (or was it Kood?) I used on this shot. Today I’m only using “Lee Filters ProGlass Cine IRND” as they deliver as good as a perfect true color result, save me a lot of dumb work in post. They are expensive, but you might, like me so far, get away with using the 100mm x 100mm versions (depending on your camera and lenses) available for less than 1/3 the price of the larger rectangular aspect “widescreen” “really for pros” versions)
Primary grade added , LOOK nodes deactivated. I just left this view in for reference, it’s not something I often watch when grading as the LOOK nodes are always on. Notice the bias towards magenta when the LOOK nodes are turned off…
…and notice how, when the timeline-level LOOK nodes are turned on, the unpleasant magenta-bias is not actually there in the intended output. This is why you grade “underneath” your LOOK, aka with your LOOK nodes on. Also of note, there are some gnarly distortions in the saturation of the greens here that I later corrected (see video below).
Even more examples
Above, some more examples (actually – accidentally because of how Apple’s pixel density thingamajig works when taking screenshots – upscaled to 4K and 5K – although the MCC “only” shoots 1080p. Would you have noticed it was upscaled and not native if I hadn’t told you?). Most shots made with various ND’ed (Kood) Pentax A110 lenses on the MCC if not stated otherwise. The close-ups of the eye was shot using a +2 diopter (Kood) attached.