Caique Silverio

AI Creative Director & Producer | 🏆 2x Runway Contest Finalist (Big Ad Top 11, Big Pitch Top 13)

[Caique Silverio @cAI_works]

I just submitted my entry to #RunwayBigPitchContest. Hoping to win the Grand Prize of $50k

"O Amor Resiste" is a Brazilian drama series set in Rio de Janeiro, following Marcos, a teenager navigating love, loss, and identity in a world that keeps telling him he shouldn't exist.

[Caique Silverio @cAI_works]

Crazy how this has 510k views on Instagram but on I have 30. Something is really wrong with my reach here 😫

But yes… Episode 4 is here 👀

And apparently one show wasn’t enough to keep me busy.
I’m also writing a completely different Halloween series that I’ll be producing in parallel with NEXT STOP

I think you guys are going to love it.
And I think many of you are going to absolutely hate it.
Which probably means I’m onto something. 👀

Anyway… enjoy Episode 4.

Made in @dreamina_ai and @LumaLabsAI
#DreaminaCPP #Seedance #luma #shortdrama #aidrama

[Caique Silverio @cAI_works]

1.2M views in instagram on episode 1 🤯
You lot turned this into a series 😭

NEXT STOP — Part 3

When I posted Part 1 I had no idea whether anyone would care about these two. Two parts later I've got people messaging me asking when the next one's coming 😆

So, thank you. Massively! The response is the reason this keeps going ♥️

Part 3 is the date. It goes well. It also goes somewhere I wasn't expecting when I started writing it.

Curious what you make of the last thirty seconds.

NEXT STOP — Part 3 🎬

Made in @dreamina_ai
#DreaminaCPP

[Caique Silverio @cAI_works]

😎 I built a wardrobe changer for AI films!

New skill I built inside @LumaLabsAI: one-click wardrobe changes for your AI film characters.

👗Scene 2 they're at work.
Scene 5 a wedding.
Scene 9 the beach. Same character every time, new outfit every time.

🧠Here's how it works:

1. Upload an image of your character.

2. Upload an image of the clothes you want them in. A reference photo, a product shot, a screenshot, anything.

3. The skill builds a full character sheet. Front, side, back, close-ups, with the face and the outfit holding perfectly consistent across all of them.

4. Need a different look for the next scene? Swap in a new outfit image and run it again. Same character, new wardrobe, still consistent.

👍🏼 It works on cartoons and animated characters too. It adapts the clothes to the drawing style instead of pasting a real photo on top.

🎬 And if you don't have a character yet, you can skip step 1. Upload just the clothes, describe the person you want, and UNI-1 generates them from scratch. It's the best model I've used for realistic people.

😎 Try my skill for free here: https://t.co/XtZDe2Jo2e

[Caique Silverio @cAI_works]

hating AI has quietly become a full personality for some of you. it's in your bio. it's your first comment on everything.

[Caique Silverio @cAI_works]

@WonderStudiosX @moonshots_pod Crazy images!!

[Caique Silverio @cAI_works]

@WonderStudiosX Damnnnn amazing guys!

[Caique Silverio @cAI_works]

wild how the loudest "delete AI forever" takes come from people who already have the budget, the crew and the career. easy to burn the ladder from the roof. 🤨

[Caique Silverio @cAI_works]

the same people who said "AI will never do hands" are now saying "AI will never do emotion." you keep moving the goalposts and calling it a principle.

[Caique Silverio @cAI_works]

My AI videos are blowing on tiktok and Linkedin and here on X I only have 24 views wtf :(

[Caique Silverio @cAI_works]

@higgsfield_ai Crazy good guys woah!!!!

[Invideo @invideoOfficial]

Watch Day 10 here:

https://t.co/c5cn8rQp57

[Invideo @invideoOfficial]

Today is Day 11 of 12 Days of Agent Two.

Agent Two isn't just Claude under the hood - we get asked this a lot. The truth is - while we use all the state of the art models out there, the Agent's real intelligence comes from its harness: a domain-specific layer built from continuous creative research that shapes how every model is actually used - more about this in the video.

But this harness is what makes Agent Two frontier intelligence for creative work - at 4x the speed, and half the price.

[Caique Silverio @cAI_works]

@umesh_ai really cool man!

[Caique Silverio @cAI_works]

@LumaLabsAI Make humans with Uni-1 and Animate with Seedance 2.5. You're welcome!

[Caique Silverio @cAI_works]

@runwayml OMG THIS IS INSANE!! Sad I saw the challenge too late!

[Caique Silverio @cAI_works]

You guys liked my little AI film WAY more than I expected… so obviously I had to make Part 2 👀

♥️ Thank you SO much to everyone who watched, commented and followed after Part 1. Seeing so many people actually enjoying these little AI films has made me even more excited to keep pushing them further.

Made with Seedance 2.5 in @dreamina_ai , generating longer 30 second scenes and then editing together my favourite moments.

🤫 300 likes and I guess I have to make Part 3 👀

#DreaminaCPP

[Caique Silverio @cAI_works]

@plasm0 @dreamina_ai This is so creative omg! I love it

[Caique Silverio @cAI_works]

I think AI film is finally here 👀

🎥Made this little show scene while testing and playing around with Seedance 2.5 in @dreamina_ai and guys… every time I use it I get more impressed with how well it understands my prompts and reference images. This is crazyyyyy 😭

🧠The workflow was actually super simple: character + setting reference images, generate a few longer 30s scenes, then edit together the best bits.

🎬I’m VERY close to starting production on my first AI short film!

🍿What do you guys think? Is it watchable enough?
#dreaminaCPP

[Caique Silverio @cAI_works]

🎥 How I made the official Wan 3.0 launch ad.

Full breakdown: the concept, the character sheets, the references, the prompts, the failed generations, and the workarounds.

A few things I go into:

Building a character sheet for every person before generating a single video, and why that turned out to be the most important step.

Designing Wan as a physical object in the film. A big purple arcade button with a light inside, because showing something this advanced as old school tech felt like a fun contrast.

Generating whole scenes in one prompt instead of shot by shot. Fewer mistakes with the room, the background and the characters, and the camera work genuinely holds up across a full scene.

The thing that changed how I work: rarely is one generation perfect, but there's almost always something usable in it. I ran the same prompt several times and pulled the best seconds from each. You don't need the perfect generation, you need a good eye for what to keep.

Plus the parts that didn't work, and what I did instead.

Honestly, the jump from the previous version is huge. The human characters look real. Real skin, real expressions, real performances you can actually direct. And they hold up across an entire film. That's a completely different thing to work with, and it's really exciting to see where this is going.

Made with Wan3.0

[Caique Silverio @cAI_works]

🤯 AI video is getting dangerously close to “just give it the brief.”

I got early access to Wan 3.0 and was invited to create an official ad for the launch.

So instead of making a normal product demo, I made an ad about a struggling ad agency with boring clients, tiny budgets and one machine that suddenly changes how they work.

The concept:

They can throw almost anything into it. A PDF brief, a website, whatever the client gives them, and turn that into a finished commercial.

After making the film, a few things really stood out:

The realism is getting weirdly good. 
Not just skin texture or sharpness. The people behave like people. Tiny expressions, reactions, awkward pauses, subtle changes in emotion. I could actually direct a performance instead of generating 50 clips and hoping someone looked vaguely concerned. And ofc 1080p generations.

Character consistency is finally becoming useful for actual filmmaking. 
I had six recurring characters, products and the same environment appearing throughout the film. Keeping all of that coherent shot to shot is usually where AI workflows start falling apart.

The camera actually understands the scene. 
In longer takes, it knows where the characters are and can naturally move its attention to whoever is speaking. With generations up to 30 seconds, you can start thinking in terms of scenes instead of individual 5-second clips.

And probably the most interesting one:

Anything → video.

Give it a document. 
Give it a URL. 
Give it the raw material a client would normally send an agency.

And it can actually interpret what it’s looking at and turn that context into video.

I absolutely don’t think this removes the need for creatives. If anything, it makes the creative decisions matter more because the execution is getting faster.

That’s the part I find most interesting: the gap between idea → brief → finished visual is getting smaller and smaller.

Workflow breakdown coming this week!