Show HN: Filmbox, physically accurate film emulation, now on Linux and Windows (videovillage.co)
We released Filmbox two years ago, and it has gotten a great response. It's been used in huge movies like "Everything Everywhere All At Once".
It's been a huge rewrite to get this working on Linux and Windows from our original Mac and Metal code.
We also have some interesting uses of cross-platform Swift + Electron in our plugin manager app, and cross-platform Swift generally in the plugin. Hopefully we can detail that in a blog post at some point.
There's a free Filmbox Lite version to try, if you're interested.
265 comments
[ 5.2 ms ] story [ 288 ms ] threadLattice being particularly close to heart. Thanks!
Have you considered creating a parallel product as an Adobe Lightroom plugin and/or a standalone app for still images?
I have a camera that shoots cinemaDNG, but I would not necessarily expect resolve to support a single dng frame. Plenty of dumb things out there, and no need to be snarky when someone learns something.
Love your website/blog... Phase deserve pinning every now and then ;)
Plus you still get Capture One for free still with them...
Yes.
It's a whole thing in the Fuji world, see http://fujixweekly.com/
I'm not sure if the xe2 is compatible but you can also edit "on-computer" by having the camera plugged in and using its on camera raw conversion but with the UI on your computer: https://fujifilm-x.com/global/support/download/software/x-ra...
For system agnostic, there's also tons of plugins, profiles, dstyles, LUTs and whatever else you'd like
https://github.com/jade-nl/dt.styles
https://marcrphoto.wordpress.com/
https://onecameraonelens.com/2021/01/12/why-darktable-is-per...
I basically don't edit photos anymore now that I'm with Fuji. It's not like I'm artistically opposed, the extra step was basically ruining photography for me, I simply don't have the time.
I still shoot jpeg + raw though because I may be lazy but I'm not insane :p
You can follow us on Twitter or Instagram for updates.
I've pretty much only ever shot film, so my opinion isn't worth much, but colour negative seems to be way more flexible than digital if all you're talking about is the capture. In terms of using film in modern workflows, digital is easier.
But I never shoot low-light scenes, and as far as I'm aware the inverse is true in circumstances where you're trying to extract as much as possible from limited light, ie. digital is more flexible when underexposure is a concern.
Besides dynamic range, that’s a real stretch to say color negative film is more flexible. That sort of film arguably does a better job of preserving highlights, but I don’t see a plausible argument for any other aspect being better. Just getting the white balance right alone…
Yes, I suppose you're likely right. And like I said my opinion probably isn't worth much.
At the end of the day they're just different, and have a lot or nothing to do with each other depending on how you frame things.
In the end, it's about faithfully recording which parts of the site got hit by photons. Digital sensors do that more accurately than any practically-sized strips of plastic with silver grains on them do. Both in terms of location (resolution) and amount (dynamic range).
It's true film handles blown highlights better, but not that much better, and digital handles underexposed shadows significantly better.
Which is quite subjective. I would say that a Mamiya 7, Fujichrome and drum scans easily match today's digital medium format. And I would say that 120 film is 'practically-sized.
4x5 transparency, good lenses and drum scanning is still unmatched in terms of overall image quality.
But there is at least one: the resolution of 4x5 film easily exceeds any commercially available digital sensor, with pretty much any halfway decent lens.
Digital sensors also require collimated light, so digital lenses correct for this, where film can produce a sharp image from simpler lenses.
Additionally, I personally prefer the colour reproduction of film.
Theoretically perhaps, but you have to look at the whole imaging system (including the inevitable loss from scanning or optical printing in the case of film). Even five or ten years ago the best digital systems were able to capture about as much detail as you could realistically squeeze out of a 4x5 negative. If you disagree, I'd be interested to see some comparison shots backing this up. I shoot 4x5 as a hobby, but I've long ago given up any illusion that it has practical resolution benefits over the best digital equipment.
It's also worth mentioning that for realistic large format photography, accurate focus and camera shake start to become significant limiting factors. The actual experience of using a 4x5 camera (outdoors at least) tends to consist of taking your best guess at accurate focus while squinting at a dim ground glass through a loupe, and then hoping that you triggered the shutter at the precise moment when the wind stopped blowing.
Finally, one has to take diffraction into account. Typical apertures for 4x5 landscape photography are around f22-32. A good chunk of the theoretical resolving ability of the film and lens is lost once you're stopped down to that extent.
Agree 100%.
To match actual resolution of something like a PhaseOne 150mp sensor with 4x5 film, you need to be getting ~55lpm on film. Which is not difficult with decent lenses at working apertures.
Scanning the resulting negative or transparency at that res (~2800ppi) without any loss is also not difficult with a drum scanner. Not everyone has a drum scanner, sure, but then the comparison is between digital sensors and consumer scanners, not sensors and film.
I've measured image detail on drum scans of 6x9 Tri-X negatives I shot handheld, and as far as I can tell 40lpm was the lower bound of detail. So, I'm confident I could produce as 4x5 colour neg/transparency that exceeds any PhaseOne sensor in terms of resolution.
I read so many posts online extolling the theoretical resolution of large format film, but these are almost never accompanied by actual comparison shots. I really think you might change you perspective on this if you tried using a 4x5 camera in realistic conditions (which it seems you have not?)
You can google for real world comparisons of 4x5 film and high res digital backs. There are very few such comparisons actually made, for all the confident pronouncements one can read in forums. Even 5-10 years ago, the resolution advantages of 4x5 were marginal at best in comparison to 50MP digital sensors.
https://luminous-landscape.com/iq180-vs-8x10/
https://petapixel.com/2020/03/19/8x10-film-vs-150mp-digital-...
Given these results on 8x10, it seems highly unlikely to me that 4x5 film could offer any significant advantage over a 150MP digital sensor in practice.
Overview - https://www.onlandscape.co.uk/2011/12/large-format-vs-digita... Test Chart Image - https://www.onlandscape.co.uk/wp-content/uploads/2011/12/mir...
Comparing the 4x5 Provia & IQ180, the digital sensor detail degrades into a blue and yellow maze of perpendicular artefacts as the lines converge, where the film degrades more gracefully (in my opinion). The digital image (in this test) appears to produce an image that is not faithful.
It's reasonable to suggest that the graceful loss of detail on film offers a not-trivial advantage in high-contrast detailed scenes (especially high-resolution black and white films). In overcast low-contrast scenes digital clearly produces better detail.
Perhaps this maze-like tendency of some digital sensors is why people look at their film and think it's sharper or provides more detail.
It would be interesting to see a test of current 150mp sensors, perhaps this has been corrected and the advantage disappears, but maybe digital artefacts like the blue-yellow maze are just moved further right.
More seriously it does depend on your definition of flexible. Being able to switch speeds (ASA numbers) without having to replace the film or carry multiple cameras is a huge flexibility win for digital all on its own.
I believe they have Adobe + Da Vinci plugins for film & video (and maybe a standalone app as well. At least they used to when I played with the beta, but it may be plugin-only now, I’m not sure).
As far as I remember, they try to do some physically accurate emulation as well. The founder is a big fan of film photography and also runs some interesting gigs (like, repackaging cinema film for use in still cameras) - also blogs about color and film photography.
https://filmulator.org
(I've been too busy on another project to dedicate too much time to it the past year, and dealing with Windows CI sucks the fun out of everything, so it hasn't been updated in a while…)
The TL;DR is that "raw footage" is sort of meaningless with video. You could make the comparison look as good or bad as you want.
If you want to continue to be obtuse, you can use the same colorspace/LUT that any generic phone camera is tuned to.
It doesn't sound that deep nor that well reasoned of an opinion. For that reason, I won't try your software nor will I recommend it to anyone.
That's not how it works lmao
Pretending image mastering and color correction is incomprehensible doesn't make it so.
It's better to think of RAW as pure data, just like infrared telescope capture data needs transformation to be displayed on a screen.
Displaying raw as-is means clipping the data to some range, which is a transformation in itself (and a pretty bad one).
Would be interested to read more about the technical details in a blog post.
Did you end up porting your Metal code to something else which is then translated into Vulkan, Metal, DX12? Or do you now maintain your original Metal code alongside ports to Vulkan and DX12? Or something else?
If you were to start another project today, would you go the same route of Metal first and then whatever else you did? Or would you go directly to the way that you are currently doing it?
Forgot to respond to this. We didn’t come up with a write-once GPU programming solution. So yeah, it’s just write it and maintain it two or three times sadly. We would probably take the same approach if we were to do another plug-in for simplicity. We went Metal first because we only made Mac software at the time.
The Avid/Nuke/Maya stack is far more popular, in that realm. Followed up by Adobe’s stack on the lower end.
on the VFX side, Maya -> Mari -> Clarisse is still overwhelmingly common unless it's a studio that has their own custom tools. Throw Houdini in there for good measure (starting to replace Clarisse more and more for lighting and scene assembly).
Consider a glass of wine. That could be described as a physical thing.
Now, consider some pictures of that glass of wine. The pictures of the glass of wine is not the glass of wine, there, we agree. But the pictures of the glass of wine is not nothing either, they are their own things.
Let's consider a fully-analog picture, it could be described by the following properties: Lens particularities, camera particularities, film particularities, development particularities, photo-paper particularties, storage particularities. Together, these make up a the analog picture. It is a thing, and those particularities are part of the identity of that thing, in addition to the subject (a glass of wine) and the other particularities surrounding the actual scene that was portrayed.
Now, change one parameter, and it's a different thing! You can change something about the scene (a different table for the glass of wine to stand on) and it becomes a different picture, obviously. You can change lens particularities, it becomes a different thing. Anything you change here, it becomes A DIFFERENT THING.
If we scan the picture, it becomes a different thing, added to it's identity is the particularities of the scanning process.. For sanitys sake let's assume we store it losslessly, exactly the bits we get from the scanner. If we have such a scanned picture, it's _NOT_ the same picture as the one we had on the paper, it's a different thing.
If we took the picture with a digital camera, changing nothing else, we still end up with a different thing, it's a picture of a glass of wine, and part of it's identity is the particularities of the lens, and the sensor and if it's a good camera it's stored in raw and nothing more is changing the information, and so that's what makes up that pictures identity.
Consider the fully analog and fully digital pictures, how may they be experienced? The analog picture could be framed and hung on a wall and viewed by the human eye, this is the only way that particular picture can be seen, if you scan it, you can see a scan of the picture, but not the picture itself. What is the true representation of the picture ? Probably there is not one, it depends on the light reflecting off of it.
The digital one can only be shown on a screen, it's identity will change if we print it, or develop it on photographic paper.. But the screen also is no true representation, it depends on the particulars of the screen.
It's the same for projection of pictures.. There is a physicality involved when a piece is captured on physical film, it's part of it's identity, along with development particularities and the particularities of the receiving medium, physical or digital. So it's simply NOT THE SAME THING.. I'm not saying it's worse or better, only that it's different, and that I know it's different, and my knowing it's different makes.. a difference.
So as I said, I love the look, but _part_ of why I love it because I know it's there for a very real and physical reason it looks like it looks.. And if that's faked, then at least that part of the love for that aesthetic is gone for me.
A physical lens distortion is significant for me not only as a purely visual aesthetic, but because my brain recognizes that this is an artifact of light going through glass. A digitally applied lens distortion may look the same all day long, but it's not real to my mind in the same sense, it's simply something added later.
Now, don't get me wrong, the simulation is _AWESOME_ and I'm happy it exists, and I'm going to play around with it.. But I do reserve the right to both like love and hate it at the same time.
Interestingly you can generate those simulations later in camera too while reviewing your images. And Darktable has some Fuji film simulation recipes builtin too for a start.
Last but not least some sites, for example https://fujixweekly.com/, publish and review various film simulation recipes.
Same problem of choice here too, and that's where the somehat tedious manual post-processing in-camera comes to the rescue, if needed.
Velvia is a slower, even more saturated slide film. Its sim colors probably read as more “unreal” to you. Fuji negative reads colder and more desaturated (to my eye), which is also probably why it may appear more manipulated and “unrealistic” in a digital context.
I’m filming family events, do you find using an actual camera changes the dynamic of collecting memories? (For example the amount of additional attention the size of the device draws?)
Also, I shoot RAW photos on iPhone pro and am very happy with the results and editing capability.
The hardest time I have is making the time to properly edit the originals to something ready to share.
It seems like dealing with another transport method from a camera would add to the friction of a workflow. Do you find that to be the case? For example, does taking on an x-systems workflow mean a greater commitment to the process of producing a finished product?
Another positive is that you can get a big camera that's virtually indestructable, unlike any phone (but then you probably had the phone anyway).
Also, yes, I find getting the photos off the camera does add friction to the workflow.
All I do when I shoot on camera is plug it into my laptop or PC and immediately upload the good jpegs to Google photos. From there I can download on phone and reupload to wherever (not Instagram, I don't care about it).
I have plans to put a section on my website for my favorite photos.
So no I don't find x workflow to require a commitment to process. It's only slightly harder than just shooting on the phone.
I disagree with others saying phone is good enough. Lens characteristics (shooting aperture priority to choose depth of field) alone is a good enough differentiator for me to justify carrying a camera.
Luckily there are very portable solutions for those that want better than phone without dropping 2k and 500 on camera and lens. Old Fuji mirrorless are all basically phenomenal and only miss out by having lower rez sensor, slower autofocus, and maybe missing one or two film sims. I shot on an xe2 with the f2 23mm for years to great effect. I hear and see great things coming from Ricoh as well which is point and shoot and very portable, not sure price though.
I think phones take phenomenal pictures but their sensors and lenses are quite tiny.
I still don't shoot RAW on my X-E4, weirdly enough.
Funny that people will spend so much time, money and effort in making digital photos look like film, but refuse to shoot film. Just sad. :(
I'd argue somewhere between no-budget and massive, there is a sweet spot to be had.
Everything else being equal, shooting on film is of course more expensive...
however film can be a forcing function, especially if you are on budget. (But not low enough you can't handle the fixed costs of film.)
It's all in the planning:
- you can't get "free" re-takes. That keeps everyone focused, including actors.
- you must have more light, so you are more likely to plan out your light, instead of shooting in different conditions and try to correct later
- film lends itself to a certain look. Shooting digital, you have so many options in postprocessing, you can lift stuff out of the shadows etc. Too much freedom can be a bad thing, film forces some coherence.
- the fact that you can't see the filmed results the same day means you get really disciplined and focused making sure everything works beforehand. When you have many people involved, this is key. (If you trust your process and use a video tap to see if scenes were nailed, you can even skip dailies to cut cost.)
- less raw video data to handle, post process, grade and cutting. The less film shot, the faster everything post shooting takes. There are many films which could have been cut into several completely different movies because there was so much material shot, and the "risk" of this being possible increases with the lure of cheap digital capture.
So to be clear, I think film can in some conditions be cheaper than digital, for social reasons. If someone has the skill to push back and keep such a tight ship on digital set, of course they will be cheaper and faster.
I base this on listening to producers saying they went for film for nostalgic or aestethic reasons, but were surprised to same or less total spend as their digital projects.
I may be full of it, feel free to shoot this all down, you are obviously in the industry and I'm just jealously peeking from the outside.
So I'm not saying film is inherently cheaper
As a non-Photographer, the film versions looked better in almost every case. I wasn’t sure what the filter was actually meant to be used for. But a comparison with neutral stock would have made it look better, in all probability.
https://www.arrirental.com/en/cameras/analog-film
Film has a long expensive feedback loop, with limited exposures, processing time, and associated costs; most users they would be better served using a digital filter.
The rule of thumb on digital is, "don't overexpose". On film, it is "don't underexpose".
If you're interested in getting most of this effect for free I highly recommend using the venerable Hald CLUT library found here: https://github.com/cedeber/hald-clut
These are lookup tables made using film scans and do a damn good job of emulating the film "feel" especially when used with HDR input without burnt highlights or crushed blacks.
I personally recommend RawTherapee for photo editing, which includes native support for these.
To further emulate the effect you can use a film grain overlay (hundreds available one google away) and a color-weighted bloom filter.
ffmpeg and OBS also natively support this LUT format, I'm sure there are ways to use them in the FOSS video editing suites as well, but I basically do everything in ffmpeg commandline nowadays so I don't have firsthand experience.
The things I mentioned will get people close, or at the very least set them on the right track. Perhaps some of them will be inspired push the envelope and contribute their research back to the commons instead of guarding it in order to collect rent.
I swear to god, this website is so infuriating.
If you would like to learn more about this topic you can look at Steve Yedlin’s discussion of the matter https://www.yedlin.net/index.html or look at some of the technical documents linked from our website.
I posted elsewhere that I thought this would be cooler if it was open source. My comment fell flat and was modded down a lot, I felt a little bad about it. Now I see this and I'm glad I posted. You post an ad for your paid software, in Show HN, then you get upset when people point out open source alternatives and criticize you for your business model. OK
We are hardly the first company to sell software, nor post about it on this website. Proprietary software is ethical.
Responding to your post is not "an attempt to silence discussion of perceived problems with our society". Such nonsense!
I don't believe this is true. I can't square creating something that can be duplicated essentially for free to provide value to people with denying that value to people in order to enrich myself.
I also take issue with the practice of software as a service specifically, and while I see that you provide a perpetual license, without pro-rating monthly payments toward a perpetual license you're creating a psychologically abusive system to extract value from people with that model same as most SaaS.
I'm not sure what you're doing is bad on net, probably the world is better off for the solution you've developed despite the aspects of it that I find problematic. I understand how hard it is to make "good" choices with regard to this stuff on a personal level, and I place most of the blame on the systems that surround us.
I think the ideas/memes I'm sharing/championing here are deeply important to the continued freedom/growth of the human race in concert with computers. I'm not trying to be mean to you or put down your product or belittle your effort. I understand probably deeper than most what it takes to do this and what you have done is impressive. That doesn't make it right. I don't have the answers, I'm not even trying to tell you "stop selling your shit and open source it immediately". I live in the real world too.
I'm very worried that most people are not even thinking about the questions/implications of artificial scarcity and how tacitly okay we are with it as a society, that's why I'm so loud about this.
But it’s so interesting to imagine a world where it’s unethical to release a film for purchase on your website, or for HBO to make an original show, or an independent journalist to publish a newsletter on Substack. Or where the only way to access music is by unnecessarily creating disposable petrochemical discs.
If we were to imagine a future society where our basic needs were met, with UBI or some other form of welfare for all (which I thoroughly support) — selling all these things would be even less unethical!
In such a world people who enjoy the physicality etc. would get petrochemical disk based music, most people would just download a file or stream. The unethical thing is denying others the right to copy and modify, the unethical thing is intellectual property.
Yes, I too support UBI but I think your conclusion is wrong, it becomes much much LESS ethical to try and deny people access to data if you don't need to make money to be alive/healthy/happy. Why try to create inequality in a situation where it's so unnecessary? What's the point?
That said, IP laws are too strong. Exclusive IP rights should probably only last about 20 years, among other changes. I think someone who was alive when Star Wars came out should be able to create their own Star Wars movie before they die. I had a very interesting and stressful conversation with David Simon, creator of The Wire on Twitter about this, where he vehemently disagreed.
Re: UBI "happy" is doing a lot of work there. It relates to the "basic" of UBI. Where's that line? IMO, an optimal target is to guarantee the approximate lifestyle of someone making perhaps $100,000 a year or so. Something comfortable but you're not buying yachts. Maybe a canoe or two. We should start at literally any number (probably somewhere around the poverty line) and raise it as politics allows as quickly as possible. I'd imagine that would take about 100 years. In that situation, I still think market effects would be useful to encourage and incentivize people to create new things and improve the world. Obviously we'd have a lot more information at that point to decide on the details!
As in, one where you'd release the source under a modified GPLv3, but any project created with it would also have to be under GPLv3 or CC-BY-SA? That'd allow your project to have a truly free software version for people who create free media projects (e.g., blender open movies).
Hardly a criticism, more just a complaint about capitalism in general it seems. Which I don't expect a software shop to be the ones to solve...
I sadly don't do anything with video editing anymore, and $129 per quarter to play with is a bit rich for me as a result, but I think what you're doing is pretty cool! I have always loved the "film look", and have found a lot of (though not all of) the film filters in the past to have a sort of "artificial" look to them. This one looks decidedly better.
(Obviously I then release everything I create under GPL, so others can use it freely and expand on it)
I support people undercutting businesses in their marketing post comment sections with Free as in beer or freedom solutions to achieve whatever's being sold. I believe in the ethicality of Free as in Freedom software, for one thing, but also in general information freedom. I feel gross about knowledge hoarding.
If you're a fan of capitalism, I think you should agree with me. Capitalist minded people can try to use Free software to extract labor for no cost (jokes on you, we extract your engineer's labor at cost to you when they get frustrated and start making commits at work), and Free software comes with other advantages. Furthermore, if you're investigating solutions to your business need to do filmic solutions, isn't it nice that there are hordes of us weirdos ready to help prevent you from spending money unnecessarily? If you're trying to sell filmic solutions, isn't it nice to be challenged on your business model publicly and specifically rather than scratch your head at poor sales with no idea how to add value?
Sidenote I was reminded of this comedy video: https://youtu.be/9kaIXkImCAM "Interview with an FFMPEG Enthusiast"
As an aside this is one of the funniest videos i've seen in a while, and totally captures the open source ux experience (i say in a loving joking way)
Now I still self-host a lot, but it's stuff that requires absolutely zero maintenance overhead. I admire people who are into self-hosting the whole stack, but it's not what I want to be doing. So I never moved off gmail, and I pay someone to host my cloud files. Life!
(Now ask me how many hours I spent getting a tiling window manager into a usable state. The number is a lot, and I feel bad about it, but now I have SUPER POWERED PRODUCTIVITY.)
/s obviously. But at least it's more like a really involved interior decorating project than reinventing power delivery to the home.
For an authentic look, halation, grain patterns, diffusion, highlight roll off, detail resolve and other imperfections are needed to complete the look, of varying qualities for each film stock. Not to mention lens choice as well. Modern lens with software correction may not look right.
Also this software is not specifically for photographers, it’s for motion picture.
I love open source too, I just don't think proprietary software is problematic.
Apparently using a free OS implies one is doing a “high-budget production” and thus demands 5x cost. Of course the “lite” version does not support Linux at all.
https://news.ycombinator.com/newsguidelines.html
(I agree that misleading edits aren't good but that's a separate issue.)
What's the value prop for a studio to run something on Linux vs Mac / Windows?
This is interesting to me to hear as the trope for Linux is that it's not used for audio/video work (and couldn't find much online about real world professional usage without pulling up reviews of best video editor / etc sort of results).
Also, congratulations on the release - rewrites are a huge undertaking.
- Brian Eno
it goes from "wtf just flickered on my screen" to "wow, that loaded really fast!
but it's an async process that could take longer, so then you get into "well if it's <200ms we won't show a spinner, but if it takes longer we will..."
it's a settings page where you're clicking toggles, and each toggle saves automatically when you hit it. so when you click one, you get a short spinner, then a check mark when it succeeds. feels right to me when I'm using it
I wrote a React hook called usePending[1] for this once. It's surprisingly difficult to get right.
[1] https://gist.github.com/cmmartti/4f7833292b9d277685fc0813532...
For example, people like 8-bit pixel art and chiptune music, but I don't think anyone misses the 10+ minute load times of the Commodore 64 (or even the 3 minute load times of the PS1). People like the goofy 90's internet with marquee and blink tags, but most people don't miss eight hundred popups happening with every website. People miss the pops and ticks and hiss of vinyl, but I don't think most people miss how expensive and risky music was pre-Spotify[1] (Youtube Music, Apple Music, etc).
I think this quote is being a bit selective with what they're choosing to remember. Yes, there are certain idiosyncrasies and limitations that people grow fond of, but I think there are just as many things that get supplanted and that rightfully get left in the history books.
[1] There were multiple times where I would end up paying upwards of $20 for an album because I liked a single on the radio, only to find that the rest of the album is completely awful. Piracy was honestly made this experience measurably better, and Spotify and its ilk basically made something superior to piracy, completely supplanting it for me.
But I don't try too hard, because there's a hell of a lot of merit in either (a) just whacking some sliders around until I hit an unexpectedly great look for the shot or (b) just letting the camera give me a JPEG that is better than just about anything I'll end up with and spending my time shooting. I still use emulation, and probably always will (and someday I'll find a good emulation of the VSCO A-series iphone filters) because of that easy headcanon situation, but the fear of getting stale is real.
In fact, for Dune they shot it digitally, copied it to film and digitized it again just to get "real" film grain as opposed to digital approximations.
The Dune example and others (Nolan films especially) are particularly annoying, so much that I literally use software to scrub the film grain off, like Topaz Video AI.