I think for now on the majority of the models that is still the case.
Not enough has been trained beyond Rec709 8bit and HD or less resolution. No clear indication if this will ever change outside of studio specific models. BorisFX and some other tools are training their models on 32bit in better conditions. But those are generally models for specific purposes, not general GenAI that we need to create shot components.
Others may have more insight to share here as well.
but at least we have Matrix Text presetā¦
This BorisFX release is more interesting then anything Iāve seen in Flame for years.
I was not aware of the new Matrix Text game-changing feature. They actually put time and dev $ into making this? Who was the one loud Flamer on Beta demanding this, and actually getting it?
While the rest of us regularly do beauty for clients who pay for that kind of thing, and could really use features like Silhouetteās in Flame.
As Iām now purchasing Silhouette for these features, adding to my Flame and other subs costs.
Curious, informal poll:
What is the most innovative and useful thing to be introduced to Flame in the last 2 years? What would you prefer the resources going into?
I know the answer is typically, āYou should be on Beta, so you can have a voice in the development,ā etc. I was on beta for years and it was great. Now running a shop, I donāt have time to track that too.
There must be a better process for choosing which features to prioritize and develop, beyond a small group of users who donāt necessarily all represent the wider user base. I canāt imagine the broad Flame community would have endorsed all this Text/Type development ā which must be one of the least used tools in all of Flame ā over begged for help with trackers, particles, 3d environments, etc.
Iād love to get a peek at the internal memos about this decision.
Given that we havenāt been able to upgrade Flame in the past 2 years since they all have horrible company ending shameful bugs, I canāt answer that question.
BorisFX are a super company.
Very responsive.
Very helpful.
Unmatched value.
Hard to not recommend them.
Very powerful and informed company.
The very thing that most people complain about:
A complete and seamless (for some of us) divorce from that stonefs poison that everyone has a love hate relationship with.
Followed by the divorce from userdb.
If youāre not exploiting those two enormous, silky smooth changes then youāre doing it wrong.
ymmv.
And probably does.
Except it is still quite broken. See my posts in the other forum.
@ALan - Except that Iām in the process of using 2027 release at scale with a lot of other people on a pretty interesting and valuable project, and the breakages are so negligibly small thats itās not really worth whining aboutā¦
So in my opinion, 2027 is phenomenally good softwareā¦
Or maybe logik-projekt is a phenomenally good way to work collaboratively at scaleā¦
Either wayā¦
Your kilometttrage may varyā¦
And it doesnāt matterā¦
I donāt have any such issues in my ecosystem, as many people donāt. Most all of us though, would appreciate innovative application features, beyond text, that are relevant to what our clients pay us for.
I think Match Grain and ML TW/Morph are two features worth noting. While neither was new, but rather more catching up to what existed, having it in platform is a big plus.
The work theyāve done on Automatte falls in the same category. As we know, BorisFX ran away with the prize on ML matting built into commercial software. So there too Flame was just catching up to stay in the game, but did it well.
There was an interesting exchange on last Sundayās Logik Live Zoom call, where Fred expanded on what they had to do to get the training data for the various ML features in Flame. It explains some of the turtle speed mode.
But beyond that, a lot of the energy has gone into platform stability - like removing framestore and moving towards file/db based projects.
Realistically it seems a capacity issue. Any platform the size of Flame requires a certain amount of energy to remain current with changes in hardware and software ecosystem (remember the move from OpenGL to Vulkan/Metal). The unfortunate reality is that with the current investment profile by ADSK, just keeping up appears to be taking 90% of available capacity. The rest then gets spend on less exiting infrastructure improvements (Text, Metadata, etc.). There was some extra speed for a year when Francis rejoined, and then ADSK did their version of AI whitewashed layoffs.
But the same is true across the industry. You have startups that try to impress, and you have some companies that have the money and risk it all - namely BorisFX and Blackmagic. Really the only two I can currently think off changing the game. The rest are all treading water or are opensource/community efforts anyway.
In fairness: if you read the forum, there were plenty of people screaming about the lack of metadata support and the lack of match grain. And complaints about text tools are as old as the industry (you should check the Avid forums). So the Flame team didnāt pursue stuff nobody asked for. They just didnāt have enough steam to other exciting and ambitious stuff at the same time.
And forums like this and all itās industry peers do create a bit of a distortion field. Well meaning and loud people freely sharing opinions. Yet, itās likely folks paying for larger license pools in whatever big shops are left, that have rightfully so more influence on what gets done. And those pipelines move slower anyway than what most of us here are doing.
Very true.
Itās horses for courses. An update to the text node would have been something I personally would have liked to have when I was finishing & mastering out of Flame. Lack of other features that were available in Resolve led to moving teams to master/finish in Resolve over Flame.
I find it interesting that Flameās ability to cover lots of things well was once its core strength but is now a weakness as there isnāt the budget to develop for all of Flameās potential use cases.
Weāre going to upgrade to 2027 for the frame based metadata toolset. Not glamorous, hardly a new feature but without it, Flame/Flare had become a real PITA for our workflow. Iād like to have used more Flame/Flare but due to highly specific metadata delivery & naming requirements, you couldnāt use them.
So, as much as certain developments may not interest some, they may be highly important to others. If Flame had Dolby Atmos and IMF/DCP mastering tools I would know several post houses that would not have switched to Resolve (or other tools) for Finishing and there would be multiple additional Flame seats to what there is today. By the same token though, if they had focussed on that market and developed those, you may have missed out on other tools instead that would be useful to someone else. Itās hard when you have a limited budget to pick what are the best things to focus on developing when you canāt get a general consensus from the user base. For instance, Alan thinks Flame beyond 2025 is broken and unusable. Phil thinks 2027 is fantastic. Iām sure they are both right in regards to the use cases they each are experiencing.
Should batch have become that? Iād love a better 3D environment. Iād love a more modern multichannel exr workflow. I was one of the people who pushed strongly for the Match Grain node and Lens toolset which have been game changers for our own use case. Thatās because of the work I am doing though but I could also see the argument for development supporting better Finishing & Mastering tools. Shame we canāt have it all but thatās the reality.
That is is the crux of it.
Our environment has become more complex with formats, color spaces, and types of workflows that exist. Being the app that pleases everyone at a time with the most stretched budgets is a thankless job. The complaints here to wit.
For my part, I master commercials by the hundreds per year, including the comp work, green screen and all the varied tricks and processes that fall under the now ubiquitous umbrella of āclean-up.ā Yes, I spent 5 years (or more?) requesting an upgrade for one of the oldest unchanged nodes in the application. I even had a previous dev team boss recommend some third party apps. I also recall that a number of people pitched far more complex solutions than what was eventually realized. I use the type and paint nodes every day and but never once in 30 years has a client ever told me my grain wasnāt right. Iām just wrapping a job with a major brand in which I had to build all the ādemosā with animating type in 4 different aspect ratios and in 3 different languages. The old text node would have been miserable to work with on such a job. ML morph and TW are killer, even though it displaced my long-revered Fluid Morpher. I donāt have enough hours with auto-matte yet to call it a fantastic time saver, but the few tests that I have run show awesome promise. I look through the new features list with every release and thereās maybe 5% that I feel I will use on a regular basis, but I know that to others, something in that 95% is an important improvement to their workflow. Who am I to question?
All very fair points. Thanks, gents.
MatchGrain and new LensDistort are great additions. was nice they added proper STmap implementation as well. I do see a lot of cool development around G Splats that would a great addition as well IMHO
Thatās not how it works. Anyone can make a feature request and anyone else can upvote it. No beta participation necessary. Most of the heavily upvoted FRās are in the software now. Sadly, the best FR is sitting at only 8 votes.