MC Galactic

Super fun creating this Betting Site ad. Which scene was your favorite?

@bojangajic94 @grok Actually the green screen scenes are ai too @grok

@MCGalacticAI @grok is this an AI-generated video?

This was CUTTING EDGE AI video last year. Today, it's easily recreatable. This took 3 days (long days) to create by one person (me.) Today I think it would take me half the time, but the results would be far better.

Real crew: weeks, locations, actors, permits. Me: a few hours and uisng leftover AI video renders. Version 2 of the AI headphones ad posted a few days ago (check it out, it's cool.) This was created in under an hour since I used left over clips. Thoughts?

AI video is moving fast. Very fast. Six months ago this was cutting edge. Took time. Took tweaking. Took patience. Back then it felt advanced. Today? You can build this in under an hour with the latest models. The gap between “impossible” and “done by lunch” is shrinking fast.

Same ad. Different cuts. Different music. Different energy. What do you think?

2025 AI. Alligator. Deodorant. No crew.

Stop the scroll. Solve pain points. Another 3 hour speed run to create a full AI video ad that would take many days and large crews and locations if done the traditional way (and big $$$.) Millions of businesses will now be able to have Superbowl level ads at affordable pricing.

2025 was when I realized nothing was off-limits creatively. Testing a fake brand, but was fun. Used at a marketing conference.

Built this in a 3-hour speed run. Hour 1: concept + script. Hour 2-3: generate, cut, score, finish. F1 driver. Firefighter. Tattoo Artist. Date Night. Bank Robbery. Surgeon. All tied together by one face mask. Not a template. Not stock footage. Full narrative arc. Three hours from blank page to final edit. That’s the new baseline.

I created this AI test ad. It is already starting to look outdated. Guess when this video was made?

A brushed metal "Kreator" water bottle Ad. To film this the old way you would need: ⚡️Giant animatronic shark ⚡️70-year-old pro surfer ⚡️Ocean crew & Film Equipment ⚡️Boats & Safety Divers ⚡️Insurance ⚡️Weeks Or… One person. One afternoon. AI video. This was created by me last year as a capability test. Today’s models would eat this alive.

How are you getting an image that you already rendered and is perfect, but you want to add the reference colors to it, to work in HEX the way you show in the video? Everytime I have used it, it changes the image completely if you try to upload your own image (the one you want color corrected.)

the new soul 2 HEX can also transfer color from another image

I just replaced a $10,000 product shoot with 1 upload just 1 product photo, it generates 9 shots while keep product, actor, color, bg super consistent and even.. edit details all on Higgsfield AI here's how:

Nice Dave! The video outputs are very good now. The way the AI puts things together is still meh though. And the fight scenes from SeeDance all look very similar (it's a tell-tale IMO.) HOWEVER... seeing this compared to where we were 6 months ago makes me believe that in 6 months from now, this same prompt will produce something that looks like it was made and edited by a studio releasing a big budget film. We are definitely getting close. And I think we are essentially (almsot) there with Human editing and a lot of renders to pick and choose the best clips for continuity.

Give me the most epic hand-to-hand combat scene taking place in a collapsing building falling sideways… Seedance 2.0: Yes

6 months ago this was crazy cutting edge. And it took a LONG time to make. Now it's not even as good as the videos we are seeing out of SeeDance 2.0 that are being created in hours.

Nanobanana with same prompts and images, 4 outputs each. Using the firery intensity color grade as a reference image.

I got a bit excited seeing the Color Hex tool from Higgsfield. The marketing ad I saw for this feature makes it look like you can drop in your image (my green cybertruck) that you wish to 'color grade', then drop in the reference image (fiery girl) that it will pull colors from, and you get something cool. Doesn't seem to work. I asked it for 4 image outputs. See attached 4 images to see what I got. If you use TEXT to image, it works fine. It's just way easier to do this in NanoBanana. Thoughts?

When using just text: "It never made an ace of spades, though, in the 4 outputs I generated." It got it there with the right colors. It never made an ace of spades though in the 4 outputs I generated. Again, just way easier to ask NanoBanana to do this, and it gets it right the first time.

@maxescu @maxescu Alex, yeah it's sometime this year where the uncanny valley goes away IMO. When you say a single prompt split into 4 outputs, how many total video renders did you have to do to get enough clips to do it all? And are these 15 second renders each?

The people who completely missed the point of this are the ones who will be left behind in this AI race. This was a single prompt, split into four outputs, then I simply stitched them together. The edit, the music, the voiceover, the story, the style, and the concept are all done by AI. Even the logo at the end. I applied a workflow that can be replicated for any brand, product, or style. All this with a model that wasn't even released yet. If you choose to nitpick about a wheel nut or a gear shift not being perfect, you're missing the point.

@CorryWiens @higgsfield_ai @higgsfield Corry, how much of the scene transitions were fully AI vs. done in post with other tools?

Seedance 2.0 is here, and it’s insane! But if I've learned anything, it's that creativity is still king. The filmmaking tools are changing. But creatives aren’t going anywhere. Created for the @higgsfield_ai Action Contest. #HiggsfieldAction @higgsfield

The fact that this was one person in one week is giga insane. Can't wait to get my hands on Seedance2 and take it for a speed run. I'll post a few of my previous 3 hour speed runs soon (with now, older tech.)