The GX85 "Super-16" project

[Clark Nikolai]

14 minutes ago, kye said:

Gate Weave increased (and I rebuilt the gate weave engine using the Camera Shake OFX).  I'm not sure I like the aesthetic, so might turn it down for my own projects.

would especially appreciate your assessments of how close this is and what might be any next steps.
I feel like it's good enough for my (untrained) eye, but if there are any more specific things you can identify then I'd be keen to keep going and see how far we can go. 

To my eye this seems like the right amount. The shot of the coffee maker seems to have noticeable vertical gate weave that I don't see in the other shots and it was distracting. Are there adjustments for vertical separate than horizontal gate weave? If so then maybe only use horizontal.

But other than that one shot it just seemed "natural" for the grainy look. By "natural" it's of course what I've come to associate with grainy film from a certain era. Later cameras were so stable that they didn't have any really.

[kye]

1 hour ago, Clark Nikolai said:

To my eye this seems like the right amount. The shot of the coffee maker seems to have noticeable vertical gate weave that I don't see in the other shots and it was distracting. Are there adjustments for vertical separate than horizontal gate weave? If so then maybe only use horizontal.

But other than that one shot it just seemed "natural" for the grainy look. By "natural" it's of course what I've come to associate with grainy film from a certain era. Later cameras were so stable that they didn't have any really.

Thanks for the feedback!

The gate weave was applied at the timeline level, so is the same settings throughout.  It's funny how it appears differently on different shots through, maybe because the coffee equipment is motionless the gate weave stands out more?

The original gate weave within the FLC looked like periodic horizontal movement, which surprised me as I would have thought that the film moving vertically at high speed and then being abruptly stopped for each frame would have resulted in more vertical positioning error than horizontal.  

Sadly it doesn't have any controls other than Amount and Rate.  I ended up completely rebuilding it using the Camera Shake OFX plugin that has lots of controls.  Sadly it seems based on a SIN function with a bit of randomness built in, so I had to play with the rate slider so it would interact chaotically with the frame-rate to appear relatively random.  It does have separate amounts of vertical and horizontal (and also rotation, which I didn't use).

I asked AI about the different characteristics and during a decent discussion it suggested:

I got it to update the numbers with a 16mm projection print and projector, which made them slightly larger, but later on in the discussions it also suggested that having a small gate weave on each frame and occasionally have a larger registration slip of 3-5px, which it said was what people remember about the look.  I don't really know how to do that in Resolve, so didn't include it.
It also suggested that a worn projector might include some horizontal wandering and that film emulation plugins might emulate that as well, which might be what the FLC was doing.

I ended up adjusting the vertical and horizontal so they were stronger than V4 but not so much that I hated it.  Another thing it mentioned was on larger screens the amount will appear much more than smaller ones, and I'm editing in semi-cinema conditions, with a FOV of someone sitting in the middle of a normal theatre.  If you're looking on a smaller device it might appear less.

When I went to see Goodfellas at a theatre that was projecting from original print distributions, they played a bunch of old ads and a bunch of previews (IIRC a lot of these were 16mm) as well as the main feature on 35mm.  The gate weave was really inconsistent, with some items having lots and others having none and being noteworthy for how absolutely rock solid they looked.

Anyway, what else separates it from looking like real film?

I suspect we're not there yet, and probably not even close...

[eatstoomuchjam]

9 hours ago, kye said:

Gate Weave increased (and I rebuilt the gate weave engine using the Camera Shake OFX).  I'm not sure I like the aesthetic, so might turn it down for my own projects.

I was going to comment on the gate weave.  For me, it's a little much.

9 hours ago, kye said:

I feel like it's good enough for my (untrained) eye, but if there are any more specific things you can identify then I'd be keen to keep going and see how far we can go.  I'm gradually disabling things in the FLC plugin and manually recreating them myself using other plugins or just manual methods.

If you told me it was shot on 70's-era 16mm, I probably wouldn't second guess you after watching it.  I think that in some shots, the blur/lack of detail might go a little too far - like the shot of all the trees with orange leaves seems a little too blurry to me and looks a little closer to super 8 than to 16mm.

Otherwise, great job!

[kye]

12 hours ago, eatstoomuchjam said:

I was going to comment on the gate weave.  For me, it's a little much.

If you told me it was shot on 70's-era 16mm, I probably wouldn't second guess you after watching it.  I think that in some shots, the blur/lack of detail might go a little too far - like the shot of all the trees with orange leaves seems a little too blurry to me and looks a little closer to super 8 than to 16mm.

Otherwise, great job!

Thanks!

The more I get into this, and the more I think about what my own associations might have been, the more I realise how variable things are.  Obviously people talk about how there isn't a single "look" for film, or even for an individual stock, as it depended on where it was developed and even who was running the machine that day, but if we zoom out then the look of film for one person might have been Hollywood movies at their local Megaplex and for someone else it might have been indy films at a tiny theatre with a super-worn projector, and for someone else it might have been the 8mm or 16mm their parents shot and projected.

Even the examples on YT are stunningly different.  I've linked to these videos before, but it's just amazing to me how different they are...

S16mm on Bolex at night (IIRC on 500T and potentially pushed a stop?)

S16 on Laowa Nanomorphs (Vision3 50D):

I can tell you, the 35mm projection of Goodfellas I saw was absolutely nowhere near how sharp and clean this is!

12 hours ago, eatstoomuchjam said:

If you told me it was shot on 70's-era 16mm, I probably wouldn't second guess you after watching it.

That's very encouraging to hear. 

I've shared this in a number of places and mostly got no feedback so I was wondering if it was more a case of "if you have nothing good to say then don't say anything", or potentially "you can't be serious, this guy is so far from the mark it's a lost cause and there's no point in me saying anything!"
I didn't think I was doing that badly, but with film emulation some people have rather exacting standards!

Assuming I've reached the goal of it not being obviously fake, I'm thinking about what I would do to grade it further, and the thing that comes to mind is a split-tone and the colour.  Unfortunately the split-tone in FLC (and IIRC the separate OFX plugin in Resolve) aren't that great, so once again it looks like I might have to go the custom route.

I'm also keen to play with the Subtractive Sat, Richness, and Bleach Bypass sliders too as I think they might be the key to getting a range of interesting looks.  Some people do film emulations and have the saturated areas super dense, whereas others don't emphasise it so much.  I remember someone posting their film emulation power grade to the LGG forums and the saturated areas were so dark I thought they were playing a practical joke on the forums, but people gave serious replies and they responded seriously too, so I just shook my head and didn't post any feedback.

[kye]

Stepping away from film emulation for a moment, the thought occurred to me to that we have softened the image enough for the 2x and 4x in-camera digital zooms to be tested.

The GX85 sensor is 4592px wide, which with its 1.1x 4K crop factor, means that the horizontal resolution for its 1x is about 4174px, with the 2x crop it's about 2087px, and the 4x is about 1043px.  Normally the 2x is usable on a 1080p timeline and the 4x is not, but with the film emulation on it, the only way to find out is to test it in the real world.

Here is the GX85 with 14-140mm lens on it at F5.6.  

Optically zoomed to about 25mm:

At 14mm and zoomed 2x in-camera:

To me this looks completely fine.  So far so good..

Optically zoomed to about 50mm:

At 14mm and zoomed 4x in-camera:

This is the big one.  I see some differences here, but not enough that if I saw this in an edit it would pull me out of it, which is really what I'd want - to know if it's usable.

For some context, I shot this test with three lenses, the 14mm F2.5 pancake lens (which is the lens I'll be using), the 12-35mm F2.8 and the 14-140mm shown above.

While the light was changing during the test, what is interesting is that the 14mm F2.5 has much less contrast.

The reason I point this out is that when comparing the 14mm F2.5 with in-camera zoom to the other lenses I think the lens rendering itself is the most significant difference.

2x comparison:

and now for the main prize, 4x:

This is an excellent result as far as I am concerned!

Caveats:

• This is with the v5 film emulation grade, which as @eatstoomuchjam pointed out might be on the softer side of the range for what 16mm film rendered
• However, this grade DID NOT include the lens emulation stack, so the lens barrel distortion, vignetting, and corner softening hasn't been applied, so that could go some way to evening out the look
• I haven't done any sharpening to compensate for any softness.  In the past I've discovered that a digital zoom of about 1.6x can be equalised by some careful sharpening in post, so sharpening isn't to be underestimated, especially in the context of an edit where (hopefully) the viewer is looking at the content and not the image
• Super-16mm film was often shot on several primes, where some would have been sharper than others, or on zooms, where the zoom would have been sharper in certain parts of its range than others, so natural variation in this stuff is potentially making it a more accurate emulation rather than detracting from it

I suspect that the 4x is on the edge of being too soft, which makes sense as it's a 1K sensor readout going through a lot of processing and rescaling, but this is a great result.

So....  the GX85 and 14mm F2.5 lens is a Super-16mm setup with a "turret" of T2.5 lenses that are equivalent to 31mm, 62mm and 124mm, and it's (just) pocketable and fits in the palm of your hand!

Happy to share the power grade if anyone wants to play with it.

[kye]

Here's a shot where I graded the 4x digital zoom to match the optically zoomed reference.  

I sharpened, added midtone detail and played with curves to counter the diffusion from the 14mm F2.5 lens.

14mm with 4x (look only)

14mm with 4x (graded to match)

Optically zoomed to 56mm (reference image)

So there we go, if I wanted to get a 4x shot to match, even between lenses, then it's completely possible.