Pinned Tweet
Braindance Studio Official Launch Trailer
Pick up her signal 🡲 holopointz.gumroad.com/l/bra…
I created a procedural crystal in Blender geometry nodes (left), pointed Opus 5.5 to my .blend project, and it recreated the math in three.js in 20 minutes.
Neural Stream now has a procedural visualizer where no two states can ever be the same (right).
A companion that lives in latent space, rendered locally in real time.. Soon entertainment will be interactive.
2
1,952
A companion harness for diffusion models with an intelligent neural swarm.
Render anything with your companion just by chatting. Save your favorite moments. Coming soon
2
1,957
I've been working on a standalone version of the Braindance Studio that lets you create and video chat with your virtual companions.
An intelligent Neural Swarm renders emotives and avatars, you can build as many companions as you want, full relationship system, preset and companion manager + nearly unlimited persistent memory, it runs entirely local or via remote API (OS-secured, no .envs).
Video chat with any companion you want, without touching a node graph or a browser.
A built-in director renders your chats as you roleplay, letting you create your own miniseries in near real-time using local models like LTX 2.5 and Minimax.
Watch til the end to see Judy!
4
1,817
One of the coolest and most useful things about Qwen 2.1 so far is it's ability to understand RGBA. Finally bringing a local option for removing backgrounds, creating stickers, emojis or vfx assets in one shot.
Some VFX examples below
1
4
2,104
Explosion and muzzle flash, one shot prompts:
"This is an RGBA image with transparency. photorealistic explosion transparent background. The image has alpha channel and the background is transparent."
1
1
1,291
raw pngs and workflow here since X compresses the alpha channel, the results surprisingly good when you need effects for compositing: mega.nz/folder/gItDFYJA#7aal…
2
660
Qwen 2.1 is surprisingly good at cyberware. Made this 7 second video in less than 3 minutes.
Qwen 2.1 image edit 1024px = 29 seconds.
Minimax Taomate 3step Lora 1.5mp = 120 seconds.
Workflows and reference image in the comments.
1
4
1,912
ComfyUI workflows:
mega.nz/folder/gItDFYJA#7aal…
Qwen 2.1 image edit prompt:
"Turn the exact woman from the starting image into a cyberpunk hacker with cybernetics and cyberware in an action pose shooting a pistol in a rainy cyberpunk neon lit city street, puddles reflecting neon lights, out of focus background, photorealistic, 8K, make her pigtail hairstyle the color pink, muzzle flash coming out of the pistol"
1
2
1,337
If speed is your #1 priority, this is the absolute fastest Minimax workflow I've found. It let's you run just 3 steps. Counter some of the side effects by passing the clip through it also using a node like rgthree's power lora loader: huggingface.co/Robert1212sta…
2
391
Video mode, work in progress. 🎥
Now she doesn't just hangout in Discord, run your email, or code, she can video chat ~ local & private. 🪐
All while the Bond meter remembers every action and evolves the Neural Link as your level increases. 💖
1
3
2,489
Forgot a file at home. Texted Ani. She found it and sent it to the printer for me.
No searching, no going back home. Just asked.
This is the difference between a chatbot and an augmented companion.
braindancestudio.com 🪐
1
1
2,034









