KnightsFan

I'll put my conspiracy hat on for a moment. Maybe BM knows about imminent competition, and want to get a few more sales out before that happens?

I saw that! What a deal--priced under the C400 even. I wouldn't be surprised if the kits come down in price to match.

Well there are certainly paths to limit AI's improvement, such as nuclear war or catastrophic asteroid impact. And eventually we will run out of resources, but a lot of poor people will die of hunger before resource limits affect AI. I don't even think legal limits will have any long term affect either. Sure if the US bans AI development and stops Google, OpenAI, etc, there will be setbacks. But other companies in other countries will step up. If you look at all of human history, technology has always improved, and always improved at an increasing rate. I don't think there is much reason to believe that the trend will suddenly stop.

So are you saying that AI generated video will never become indistinguishable from real video? Its still obvious in most cases right now, my point as that it will become better. CGI is difficult to keep improving because it's largely a manual process. Some CG artists are very good, but many are not. It took decades for us to go from moving a cube across a screen to the very best CGI today, which takes enormous effort from large teams of extremely experienced and skilled people. AI generated video has gone from nothing to this ad in what, two years? And how many people were involved? And the same AI model will continue improving, and never retire. Even without improvements to the methods of training or reinforcement learning, imagine the output when it has 10x or 100x the training data.

*yet It's hard for me to believe that 20 years from now, AI will be unable to produce videos of any type that are indistinguishable from videos shot with cameras. It's an interesting discussion how much diverse training needs to be given to a model before its output is presumed original. I mean, we're all trained on images from other artists, and at the moment the primary difference is that our brains are way less adept at permanently storing precise information. And at this point, these models have ingested more videos than any of us have seen, so the inspiration is even more spread out than a human's. It's probably not that far off before some of these big companies start collecting training data straight from the real world, that combines binocular vision, audio, and other sensors into a combined stream, like our own.

1) With the caveat that I don't use AF, the Panasonic S5 is a wildly good value at $800 used (what I paid for mine). Imo, the improvement in noise and dynamic range over the GH5 is worth it, but if you're always in well-lit areas maybe you won't notice much benefit. If you aren't tied to existing lenses, camcorders can be more convenient than hybrids. I'm not familiar with available models since I don't shoot events, but I'd research camcorders for event videography with no photos. 2) C400 or FX6. The C300 Mk III is also fine--I choose full frame because of the lenses I own. It also depends on how much weight matters, as the FX6 is extremely light. If I need photos as well, then add a photo body in addition, and honestly everything is so good these days that I'd take anything modern. 3) Assuming 1-3 person camera crew, it depends on the script. My general pick right now for myself is a C500 Mk II (cheaper than a C400). I used to have Z Cam, which has unbeatable wireless monitoring and control, but I don't like the lack of color management support in Resolve. I would pick a Z Cam F6 if I need to fit the camera into tight spaces. If it's a lot of fast action like a car chase, a Komodo's global shutter would be great. The Ronin 4D is also very interesting for extensive handheld--never used one myself, though. If I can step up in budget, I'd love to try the Blackmagic 12k LF.

Agreed, that's why I don't think they'll update the S1H unless they have a new sensor. But it's always going to be the case that more specialized models will command a much higher price and have far fewer sales.

The obvious one is support for their 32 bit XLR module. But I agree that more features cost money, and people who expect a revolutionary step forward for $3k are going to be disappointed. I tend to think that the S1H successor will have a different sensor with faster readout and higher frame rates. Either that, or they will need to add non-image features like NDs, SDI, builtin XLR--and while entirely possible, I don't think Panasonic will move in that direction again.

I definitely agree that more SDR luts would be great. It would be nice if cameras could share luts with the same ease as using a new snapchat filter. Or if you could connect your camera to an ipad wirelessly and build the lut in real time using color correction software. There are so many UI/UX barriers to cameras still. However, for narrative projects, I would never shoot anything other than a standard log profile anymore. I shoot in log on large projects because it allows a color managed workflow. If I shoot a scene one day, and then have pickups a month later, with slightly different lighting because it's cloudier--or with a different camera system altogether--log is very useful. Partly because of added dynamic range, but, more importantly, the shots are standardized to reduce the guesswork for adequately matching. I also shoot a kelvin white balance rather than balancing to a card for the same reason. Standardization is key.

Google's version looks good, too. I do think we're still a couple years off from good, reliable video generators for serious videos. 2025 will see a flood of "content creators" using it for sure, but gluing together multiple layers of neural nets and traditional programming into a cohesive unit will take some time. Image generators typically work on multiple abstraction layers, so the model has a concept of what a "cat" looks like inside a prompt "cat holding a beer." To solve physics and object permanence, I believe that video will need to have a concept of 3D space and objects in that space, and specific characters are conceptualized as a character that can be reused, etc. So I think significantly more work will need to be done on each layer of the network to get beyond making portraits with minor movement. Then of course going off my earlier comment, I think that a significant constraint for something like Sora is that it's designed to be used by absolute amateurs. Professional software, like an integration into Adobe CC or DaVinci Resolve, can expose more controls or even basic scripting (e.g. in Fusion) and expect users to reference a manual to learn it all. The user base for that is so much smaller, it will take more time to get there.

Which is why I mentioned Z6 III. For the R5, if you're comparing used prices, then I'll revise to "I won't complain about Panasonic cameras until Sony, Nikon, and Canon all have real next-gen FF sensors in $900 used bodies" The Nikon Z6 came out in 2018, and the S1 in 2019. Z6 III released this year, so if a Panasonic camera comes out with a similar sensor, then next year will have the same latency in terms of time-to-new-sensor.

I don't think Panasonic will release another ~24MP FF flagship until they have a new sensor, which they probably do not have access to for a price that makes sense at this time. This "release" is simply continuing an existing product under new regulations. Lots of companies are in the same spot. Nikon did something sort of unique with partial stacking in the Z6III, but there really haven't been any new, affordable FF sensors recently. Obviously, for people who need the better autofocus and dislike the S5II size/weight/whatever, it kinda sucks not to have a slight update to the S1, but that's got to be a tiny, tiny slice of users. Personally, I won't complain about Panasonic cameras until Sony, Nikon, and Canon all have real next-gen FF sensors in $2500 bodies. Lenses, on the other hand... I wish L mount was a little more expansive and creative! But it's okay, I'm still 100% EF mount, and my adapter works fine.

The A7c II is one of the few cameras that I would rather have than the S5 I bought, but at >2x the price (used vs. used), it wasn't worth it. Even A7c used prices are a few hundred more than what I paid for an S5! And I'm way happier with S5 video than A7c.

It will be especially interesting if the "competitor" software rely on the same underlying model, or whether anyone will make their own. These models often work in layers, where lower abstraction levels are still using just a few models that may or may not be updated or improved upon. My prediction is we end up with complex AI tools that end up like this classic xkcd, where the critical pillar will be a single image recognition model that a couple grad students trained in 2008 with 1,000 images they took around campus. https://xkcd.com/2347/

Really impressive. I think what's fascinating right now is the nice user-friendly buttons and sliders like "remix strength" that they have. Obviously OpenAI has designed this to be usable to the general public with very broad controls. I haven't looked into any API for Sora, but other platforms might be able to expose more technical or complex controls for more advanced users. Same applies for the filters they have added. If I recall, ChatGPT refuses to talk about certain subjects, whereas the API has no such limitation.