I’d like to see a high quality implementation of what Finn is talking about. This was the type of industry leading features that made Flame king back in the day. Just because other tools don’t do a great job doesn’t mean that something isn’t worth aiming for which is what Finn is saying. A good implementation of an AI infill tool that is completely localised would be a valuable tool, especially utilised with the Flame version of CopyCat. You can find some massive time savings with these tools combined. Apologies as I can’t remember the name of the tool or who developed it for Flame it as we have Nuke so just use CopyCat.
This kind of feature request doesn’t necessarily mean there aren’t other features that he may also like to see.
From my perspective, it’s hard to imagine Autodesk beating the likes of something like Omni or Seedance for that kind of work. I think it’s about official frameworks that allow to integrate those API keys and harnesses to control them in ap alongside pipeline-esque modalities for making the results feel like an integrated piece of the kit. At the end of the day that’s ultimately what Adobe is doing… but with shittier models
i mean didnt they buy some AI VFX company that did inpaint/outpaint and even matchmove ? what happend to that?
Couldnt care less if its some external API , internal models, whatever but I need a “extend to 9:16” button that actually works the same or rather better than all the free stuff out there.
i think @AdamArcher said it well, i want industry leading, breaking news type stuff that make flame the absolute no brainer, right now its role in the whole process of commercials finishing is getting less and less and less everyday… as man i am not going to sit there and paint a sky extension for 2 hours in batch, these times are just game over. so i will start in comfy , and then throw in the final result into the flame timeline, at that point I might as well jsut dont bother with using flame for manyyyy things,
bread and butter type stuff not special VFX heavy whatever
The industry leading tech isn’t being made by any company making software for our market. For those companies it’s a given that we’re already dead and not worth worrying about.
Not to go all scorched earth or anything. It’s just the reality of why we need to build frameworks.
I’ve been saying this for awhile, it isn’t about AI making VFX better/easier, it’s about going from Script → Final Picture in one step. Hollywood is going to be cooked, but Advertising is the appetizer on the plate coming out of the kitchen right now.
only grading stays because they need a reason to order sushi and its faster than trying to tell a AI what they want because they dont know what they want.
I am not kidding when I say i am looking for alternatives of what I will be doing the next 35 years… i dont think it will have much to do with media and creative stuff, maybe something where I dont have to deal with assholes as much as well
I don’t think there is any solve for the asshole factor.
But for the rest - it’s important to separate what you do functionally, and what you bring to the table as person (experience, maturity, soft skills).
Depending on how I count I’m on my 2nd or 3rd career at this point. I’ve worked in very different industries and markets. There is always something to do and make money with. You just have to be curious and keep moving.
With that in mind, it really doesn’t matter how ads will be made in 20 years, and what entertainment we’ll consume in 20 years. People have worked in entertainment for hundreds of years. And we still do.
Be kind, be knowledgeable, be helpful, stay curious. You’ll be alright.
It does entertain some people to a degree. But when you overdose you can kill the golden goose of infinite scroll media. The theory of Sora’s untimely death was that it wasn’t just the price, but also a threat to the golden goose.
But that said, I grew up in the off theaters of Berlin in the 80s. Without a TV and the Hollywood slop playing in the multiplex was frowned upon by the parental units.
I would consider the Hallmark channel slop as well, just classic slop at this point. Many DPs have at some point in time shot a Hallmark movie to pay the mortgage, but rarely talk about it.
Albert Watson is known for his Vogue covers and editorials. Yet he directed 600+ TVCs as well.
When I left tech and moved into the media industry, all my dreams were about the cool stories and images. It took me a while to understand that you have to produce some amount of slop to earn your spurs and pay the bills. You never talk about it. But if the cool stuff makes up 10% of your pipeline, you’re ok. A few eventually do a higher percentage. But anyone telling you they never do slop, even at the short peak of their career, I would say they’re stretching the truth.
I hate it, I need it, i dont want it i would hope it would just all go away, i have 0% passion for any of this slop, and I guess so are the developers of VFX software.
but there I am , i need to stay competivie in the market , and pick my battles with projects actually worth spending time on vs slopfest.
I cant stop this stuff, I couldnt stop vertical video taking over and I still hate it, i am just grumpy at this point that this stuff just hasnt developed in a direction that makes me happy and there is no sort of sun at the horizon… its just like look all the stuff you know how to do and have a passion for i can now do on my phone lol
I want to spend my time doing cool stuff that entertains me, an yea even slop films had fun tasks now its all “just prompt it yolo” and the measurment of technical quality has gone down to shits.
and yes , absolutely there are still some amazing projects out there, its just getting less in terms of percentage
Sorry, I don’t buy the AI is going to take all our jobs mantra. There has been a very real pushback against AI generated advertising as the uncanny valley effect reduces authenticity of a brand. It has been researched and documented and is a real thing.
Not sure if any of you have got teenagers/kids? They HATE AI generated anything. It is a sure fire way to get what is one of the most targeted markets for advertising and media to not want to look at your stuff. Then there will be a whole load of passionate movie and series viewers who won’t want a bar of fully AI generated content no matter how good it is
It still take humans to craft it even when it is AI generated and most of the people I know working with it says the time and costs aren’t that different. If what you are making needs to resonate with people then it needs people who understand the human perspective.
Doesn’t mean that we shouldn’t use AI tools that help to streamline processes where it makes sense, such as the use cases @finnjaeger has suggested, but who in their right mind thinks we should just give up because AI is going to take our jobs?! With that attitude you probably will lose your job as your heart obviously isn’t in it. This is just another evolution in the post & VFX industry, such as moving from linear online to non linear, From practical visual effects to CGI, From film to digital acquisition. From SD to HD to UHD. From SDR to HDR. From 2D to stereoscopic right back to 2D again. At least Finn’s suggestions are trying to evolve along with the industry. Flame has been good at that before so not sure why it can’t in the future.
Oh, and for those suggesting we plug into external tools, I’m all for it. We’ve had sparks in the past and there’s OFX now as well as ONNX support. You just need to have the marker that 3rd party software providers want to write for Flame. Maybe there needs to be a new standard created for this as a modern toolkit? I don’t know but if there is a market then someone is going to try and monetise what they are making if the demand is there.
Oh, and don’t get me started on the security vulnerabilities that a lot of the GenAI tools have. I hope anyone implementing them has done their research. A quick Google on it will show up some horror security stories. There is no frigging way I’d be exposing a business to it unless it was in an air-gapped/sandboxed environment. You can’t load your content to the cloud for any of the work I am working on either so no way would go there if that is a suggestion.
Also, what about getting rid of this inference node?. I haven’t found a way to not having the model saved in each batch iteration. If I have a project that saves in 5 seconds and add one of these, it then takes a minute to save. The batch setups folder becomes gigantic.
For good reasons, we centralize all our models so we don’t have duplicates, but Flame does the opposite: it duplicates them as many times as possible, which makes the tool unusable and forbiden.
Speaking as someone who has tried and repeatedly been disappointed by AI tools, 85% is not good enough. And as many have pointed out here over the years, the difference between 70 or even 85% and “acceptable for final” is often the difference between minutes and days. Your clients want fast solutions quickly, that’s fine, all of ours do, but one of the enduring responsibilities of flame artists is being able to educate our clients to the difficulties of taking a quick one button fix to the level that they’re actually paying for.
I’m lucky to work somewhere that tries to maintain a standard of excellence, so those conversations usually go pretty well, but I feel for folks constantly inundated with “I did this in five minutes, what’s the problem?”
But we got the big chair, and we can either have the tough conversation and then put in the work, or, I dunno, get better clients.
On another project I’m working on I’ve compiled my OFX against the installed onyx version shipped with flame to run inference on another model I want to work with defined by a simple path. So… that’s an easy enough problem to solve. A single simple node that runs the inference for a centrally pathed model. Don’t think it will be super performant but it’s working for me.
