Welcome to our community

Be a part of something great, join today!

Register Log in
  • Hey all, just changed over the backend after 15 years I figured time to give it a bit of an update, its probably gonna be a bit weird for most of you and i am sure there is a few bugs to work out but it should kinda work the same as before... hopefully :)

Gemini Omni vs Seedance 2.0 vs Kling 3.0 — Ultimate AI Video Showdown (45 Videos, Who Wins?)

They each seem to be better at some things than others.
 
Just heard a producer I did a film with is considering making a "coming of age" film with generative AI.
Director who works with that producer was gauging my interest in what I could only imagine performing "cinematography" on the prompts to the AI engine.

I wasn't interested.

While I think generative AI can be a powerful tool for filmmaking I can see where this can fly off the tracks as far as humanity is concerned.

I can only speculate but I think reducing visual expression and craft to AI prompts will do to filmmaking what automatic spell checker did to spelling.

There is a balance with using AI along with traditional workflows but I wonder how much damage and resulting "scar tissue" will result with the incoming abuse of this tech.

I think long term use of this AI will be a considerable step in the "dumbassification" of the common person in a way never seen by humanity.

BTW- there are some influential "stakeholders" out there who believe the world is overpopulated with humans and something needs to be done about it. I leave to anyone reading this to make their own connections.

Brian Timmons
BRITIM/MEDIA
 
Last edited:
I'm on another forum, Colourlab.ai, which I am a beta tester for. The newest version which hasn't been released yet will now, with a few other NLEs, be using almost exclusively Agentic Ai workflows. This means that there will be very little the user of the program will be able to modify after the AI makes a decision on a grade which will be based on a REFERENCE STILL image of a past grade or Still from Shotdeck or other Website that offers stills from movies.

Before with the correct version of Colourlab.ai , the AI would allow more input for user modifications of the AI Match of a Reference Still. Now most of the tools that were once present for additional modifications of the AI grade have been removed due to company's belief that the updated AI model used is far more accurate in Reference Still matching. Another downside of all of the new Agentic AI workflows from all of the companies that will be using it, is that it requires TOKENS for certain levels of matching. Some Tokens are provided every month by the different companies but after the Tokens are used for certain Agentic AI tasks for certain durations of projects and grading, more will have to be purchased for additional Agentic Ai tasks. If you are doing multiple long form projects with multiple client redos etc, that can possibly get expensive for something that before these Agentic AI workflows came along used to be free.

I'm not to keen on turning over all the knowledge that I and I'm sure many others have gained over the years to an Ai Agent. Plus those individuals more familiar with using Ai based tools will eventually be much better than most of us whom are not that good with Agentic Ai workflows in certain aspects of Color Grading,Editing..etc whom don't have maybe 1/10th of the experience we have.

I still mainly use Resolve Ofx, DCTLs and other included Resolve tools mostly still but this Agentic AI workflow thing is seeming the future that I'm not so sure I really want to be part of if I'm honest.
 
Last edited:
There are obviously different expectations about what output AI should generate. For one, there is the end-to-end approach that gives us a finished product, picture, movie, song from a prompt. This is the consumer-oriented view, with little possibility to modify the result except by re-prompting, which then again runs through the whole AI pipeline and produces a slightly or even very different result.

And then there is the other approach that AI should make a skeleton output that is not the final result but create all the settings or objects that are used somewhere along the pipeline. Like create the required nodes for this particular look in Resolve or a complete Fusion comp for an animation or transition. And another but different AI task that mixes audio in Fairlight aligned with the video, places markers in the timeline etc. that are helpful for a human to refine the output further. More like AI creating the source code for an app rather than releasing the published app in a store. That’s the professionals’ expectation from AI and that’s what I see Blackmagic doing now with all the different AI puzzle pieces in Resolve, not many others so far.

But this approach is much more flexible and effective for anyone who needs professional results nowadays because some tweaks by a human can eventually be done faster and avoids the often unpredictable results from re-prompting. And it could go even further, for example, if AI wouldn’t create the finished pictures but create 3D models and animations of everything in a scene, which can be modified further by a human before it is sent to a renderer.

But apparently every AI company is only looking at the consumer approach these days, probably because they see the mass market in this field. Whether this is also where to get the most money, they will have to find out.
 
I've been trying out CLAUDE using the "SONNET 4.6" Model, I think "OPUS 4.7" might do a better job however. But it not very good with complex grading work. In the example below I use a grade that I have made that I have made for one image for CLAUDE to apply to another Image with similar lightening. I have already added the grade from the previous Image to the second image used so I know what it should look like when applied.


UNGRADED FIRST IMAGE



GRADED FIRST IMAGE



UNGRADE SECOND IMAGE


GRADED SECOND IMAGE with Same Grade as First Image. This should be how it will look if CLAUDE matched the first Image's Grade to the second image successfully

Prompt I gave CLAUDE




CLAUDES RESULT and Explanation of what it did


FINAL RESULTS. First the SECOND IMAGE GRADED with same grade as the First IMAGE then with CLAUDE


 
You must log in or register to reply here.
Back
Top