Leaderboard

My mind keeps coming back to the GX85, and I've realised that (yet again) it's far more capable than I'd given it credit for. To this end, I did some additional testing by shooting some night cinema on Halloween. This batch has my Super-16mm film emulation on them, but there's other grades below. What I've worked out is that there's an incredible synergy between the: The GX85 itself (of course!) paired with small lenses Resolve and the Film Look Creator The 1x, 2x, and 4x in-camera digital zoom function It will come as no surprise to anyone here that the GX85 combined with small lenses provides a very small and light rig, the tilt screen and customisable buttons gives great ergonomics, and the assist features are great too (setting the display to B&W, have coloured focus peaking enabled, exposure meter and the histogram all combine to give a great UI that doesn't need an external monitor). Where the synergy comes in is using the in-camera crop modes combined with the Film Look Creator in Resolve. If you like the look of Super-16mm film, then the 1x and 2x and 4x modes are all usable and give the following FOVs with my tiny lenses: Panasonic 14mm F2.5 AF lens gives 31mm at 1x, 62mm at 2x, and 124mm at 4x Panasonic 25mm F1.7 AF lens gives 55mm / 110mm / 220mm FOVs TTartisans 17mm F1.4 MF lens is faster than the above and gives 37mm / 75mm / 150mm FOVs Risespray 35mm F1.6 MF lens is fast too and gives 77mm / 154mm / 300mm FOVs If you prefer the extra resolution of Super-35mm film then you're still good with the 1x and 2x modes, but the 4x might be a little too soft depending on your tastes. It does depend on the lens though, and lots of stuff shot on S35 in the past was on lenses that weren't as sharp as the GX85 4x mode. Having the combo of a 31mm / 62mm FOVs, or 37mm / 75mm can be incredible for shooting events and street stuff, not to mention travel and general out and about shooting. It's even quicker than using a zoom because you don't even need to re-focus when swapping FOVs! Here's the 1x, 2x and 4x on the 25mm with my custom S16 emulation which also includes a lens emulation too: If those are too retro / soft / degraded, these are with a 35mm grade and without the lens emulation, so should be a bit cleaner: Or you can make it cleaner again by lowering the diffusion a touch and adding a tiny bit of sharpening under the emulation: and here's the 1x / 2x / 4x with that grade: Hell, you could even use the 65mm setting, which when combined with some sharpening looks like this: I'm not sure if it's going to replace my GH7 as my main camera, but in situations when I don't need the extra DR or IBIS modes, it's a real contender for sure. The level of enthusiasm about the new GoPro shows there's interest in smaller camera packages and having used the setup again I'm reminded why - the experience is so freeing. The GoPro has advantages over the GX85 for sure, but for this type of shooting, the GX85 with its IBIS wins hands-down for me. Also, the better at grading I get, the less I miss the OG BMPCC and the more I feel the GX85 does what I want the OG BMPCC to do.

With the help of our AI friends, the first Panasonic hack since the GH2 has arrived. So far the S9 and GX80 have the hack. It auto-executes code from the card at boot, enabling the same kind of range of mods Magic Lantern does but on Lumix cameras for the first time. It takes compiled C code binaries and Arm64 code. The CPU is extremely powerful... 64bit vs the 32bit EOS R5. More info soon. GM5 and GX9 to follow

Are you just testing the ability to hack and moving on, or will you eventually add features like raw video, log profiles, etc... where possible?

Solved! Live View over Micro HDMI on Samsung NX2000 is finally possible Hey everyone! For years the consensus was that the NX2000 hardware physically blocked Live View when an HDMI cable was plugged in, forcing it into gallery mode. We managed to bypass this limitation by using a hidden factory 3D demo mode (hdmidemo.adj) and applying in-memory RAM patches to force it into 2D, redirecting the native 720x480 live buffer directly to the TV video window via /dev/fb0 ioctls. Continuous AF/MF switches and the lens ring focus are working live too! You can check the full source code, scripts, and how to set it up here: https://github.com/michumichifu/nx2000-hdmi-live

Local inference on 12GB VRAM, alpha mattes, 16-bit EXR, ACEScg and 3D-guided rendering are a different proposition from generating random clips from text prompts. If temporal consistency, edge quality and render times hold up in production, this could significantly change what smaller VFX teams can achieve without expensive infrastructure. The real question is no longer whether AI has a place in filmmaking, but how quickly production workflows will adapt to it.

I have seen that, thanks. Idk, how much more recording time you'll get shooting 2:1? It's my understanding that the camera can only give an estimate based on uncompressed cDNG not with lossless. Right now with OG3.5K on a 128gb card, the camera shows 6-7 minutes. Google AI told me, I could expect 10-11 with lossless, which is great but still not enough to justify, considering how many V90 SD cards I'd have to buy. The OG2.5K is so good that 30 minutes (roughly in lossless) to a V60 card is perfect for my needs. Don't get me wrong, I prefer the 3.5K, and would consider it with an SSD, I suppose, if I look at it like Hitchcock would... shoot VistaVision with a 1.85:1 release. I also have to research, test, if the hack is on a card, can I just leave it there and still record to SSD. I haven't tested it enough to the point where I needed to format cards, but having to delete clips, or upload the scripts every time is kinda annoying. All that said, I have been swamped with work and family stuff, so I haven't had much time to mess with it. So far, just my first half hour outing, but this has completely changed my entire gameplay going forward and I have been rearranging some buying/selling of lenses, and working on some scripts to utilize it. If anyone is interested, here's another frame from the OG2K mode with the Minolta 35mm 1.8... the other two I posted were at iso 3200 and I think this one was 100... sorry I couldn't find a cat to shoot...

Maybe becoming used to low effort software has also taught people to be skeptical when the real thing shows up. As you said, pros used to load ML on the camera they used for a real income, myself included, but there's no way I'm putting someone else's vibecoded mod on my P6k Pro. I guess ML came (comes?) with a comminuty that help troubleshoot and support which felt exciting to be part of. I did however do that hardware mod for my P4K, which puts a V-mount on the back and moves the monitor to the top. It was made by someone who put in real knowledge and thought, and that's what's worth paying for. But of course it was a lot more effort on my side than simply loading up an SD card. Last thing on the Switch comment - the software used to play them on Mac doesn't supply any keys or anything, they all come with the game when you buy it. You could equate it to burning CDs. If you own the original CD (or in this case, game) you can make a usable backup. It's only when you start distributing or illegally obtaining that music that problems could occur. Whether people actually buy the Switch games or find them elsewhere isn't in the NXNative dev's control, so how can they be liable?

I expect you still need the private keys to break encryption so they run without the Switch's ROM's encoder/decryption service. This is nothing like how Magic Lantern worked. You should be grateful there are people willing to take that risk. And at least they make the effort to understand the legal side, unlike the armchair internet experts. Yes I agree, how dare they run a business in communist camera town. ---- The same people whinging about an indie dev selling a genuinely amazing tool for $5 are the ones who happily pay Adobe $50 per month on a subscription, the monster which their dumb attitude towards software licenses created in the first place. I have never seen such an entitled generation of users in my life when it comes to software. Every smartphone app is now freemium junk, that relies on gating features behind a subscription because nobody wants to pay the $5 one off up front. Open source and free is the fastest way to find the bottom of the software quality barrel. If you can't pay the devs, yet expect them to answer customer support requests 100 times a day, how do you expect them to keep grinding at the millstone and updating the project? Why do you think so much of the open source stuff gets abandoned by the main devs? Even Magic Lantern at one point too? I just think the whole attitude is fucking stupid to be honest.

Have you seen the updates on the site here? https://www.elcacharrista.com/en/articles/hacking-the-sigma-fp-with-ai/ New updates 28 October 2026 · composer on Bei’s Loader v4, with Res-Custom 0.1.6j · 2:1 3.5K format (3552×1776) up to 50 fps, framed 2:1 on screen, and 12-bit to SD. 5 October 2026 · fpSup-Formats v0.6.0 · new OG4K format, Crop Mode on OG4K and OG3.5K, lossless compression and gyroscope data that remember their state on the card, and our own composer to build it. 3 October 2026 · fpSup-Formats v0.3.2a · S16 up to 100 fps.

Runs locally on 12GB VRAM... https://ltx.io/solution/vfx