Joseph

AI Software Engineer and Content Creator. Making apps, music, films and games.

[Joseph @jjoseph_serve]

Same AI model. Different API. Very different price.

I've been building EverScene 🎬 — a tool to compare AI creative API pricing across providers.

🔎 Track pricing across 44+ model configurations and 29 providers
🎥 Compare video, image, music, and voice models
💰 Estimate costs for your actual workload
⚖️ Compare providers using the same model, resolution, and settings

Less time comparing prices. More time creating.

Check it out 👇
https://t.co/PlxhErlU2E

Built for AI filmmakers, creators, and developers. Would love your feedback.

#AIFilm #GenerativeAI

[Joseph @jjoseph_serve]

@matchaman11 Nice. I'm thinking of just using grok bots to work as a content management team - they can even use my local browser to post to the socials. Would that work too? Any gaps?

[Joseph @jjoseph_serve]

@Tiny_upgrade I like that approach. I’m experimenting with motion boards myself -- more about mapping out what the characters actually do than locking in every frame. Feels like a good middle ground between giving the model direction and letting it be creative.

[Joseph @jjoseph_serve]

Are we still storyboarding AI films the old way?

I’m experimenting with something different: a motion board.

Instead of designing every frame, I’m mapping out the action — where Nova moves, how she crosses the terrain, and what the Zergling does next.

The goal is to give Seedance 2.5 enough direction to understand the scene, without boxing it into my exact vision of every shot.

Maybe the job isn’t to tell AI what every frame should look like anymore. Maybe it’s to show it what needs to happen and let it figure out the rest.

Curious how other AI filmmakers are approaching this. Do you feed Seedance a storyboard, a motion board, or just a prompt and let it run? 🎬

#AIFilm #Seedance

[Joseph @jjoseph_serve]

I compared Atlas Cloud, Revana, and Sume on what AI video and image generations actually cost.

No single winner: resolution, developer tiers, and stale pricing data change the answer more than you would think.

Full breakdown 👇 https://t.co/RT82q18eqF

[Joseph @jjoseph_serve]

Shot 2 🎬

We introduce our heroine — Nova watches as the Zerg patrol moves deeper into the perimeter. One wrong move, and her cover is blown.

The Hive is getting closer. 👀

#StarCraft #AIFilm #Cinematic

[Joseph @jjoseph_serve]

Like Ben Affleck, I spent a couple of years deep into StarCraft in my youth. An absolutely brilliant RTS. 🚀

Now I’m building a 3-minute AI short film inspired by StarCraft to put my new filmmaking pipeline to the test.

First up: establishing the Hive. I generated these with five different image models.

Which one would you pick — and why? 👇

#StarCraft

[Joseph @jjoseph_serve]

@JanStigz The Zerg Brood is basically a giant living organism that takes over the planet. The Hive is its heart, surrounded by creepy, fleshy structures that grow and spread. Think of it as the planet itself turning into a monster.

[Joseph @jjoseph_serve]

@jadenchX Yeah, that's my favorite too. I like the sky and dark forboding feeling.

[Joseph @jjoseph_serve]

@grokbotstudios Most likely yes. I'll post a couple more stills and you can see the differences.

[Joseph @jjoseph_serve]

@peterpme What is a software factory? Not following what you mean.

[Joseph @jjoseph_serve]

@andyorsow Haven’t tried this yet, but seems like double the work. Why not just use a still, or storyboard and good prompting.

[Joseph @jjoseph_serve]

Upload both pictures in one message, in this order:

First, upload the shot you want to change.

Second, upload the picture that has the look you want.

Then write: "Change the first picture. Keep the same person, the same pose, and the same background.

From the second picture, use only the colors, the materials, and the light. Worn metal, hard daylight."

The first picture is the one that gets edited. The second picture is only a sample of the look.

If the background changes, add "leave the background as it is in the first picture."

Use one photo of the person. Two photos of the same person usually turn into two people.

[Joseph @jjoseph_serve]

I'm realizing how important it is to get your stills correct before rendering any video. There's also many models to choose from, but I'm limiting the models to those that can take multiple reference images as inputs. The side-by-side is very interesting - I think sunburst has the best take in this example.

[Joseph @jjoseph_serve]

The pipeline also tells you how stuff will cost to generate - for example I need to generate 77 still images and that will cost around $6. I'm currently using GPT sunburst for the images as I need a good model that can handle multiple reference images.

[Veee @vikktorrrre]

Jensen Huang: We can’t design a chip anymore without AI.

“At night, our AIs are exploring design spaces vast and wide that we would never do ourselves.”

“Our chips are so much better because of AI. We can reduce energy use and get higher performance.”

“Our software, we can’t write software without AI anymore.”

“We use AI to find bugs, figure out what’s wrong with the code, and who’s likely involved.”

“I want every organization and company to use AI very aggressively.”

“I want to turn NVIDIA into one giant AI.”