
⚠️ profile moved to @hipnopol ⚠️ Artificial dreams and oneiric links.
Joined November 2025
- Tweets379
- Following210
- Followers34
- Likes1.4K
Friendly reminder for my dear few followers, that I'm switching accounts → @hipnopol
2
32
I'm moving back to my old account, because there's too much gems. and Today I ACTUALLY CAN make it on my own 🤯
1
36
profile moved → @hipnopol retweeted
Published my interaction lab so you can play with it and get some inspiration from it!
It contains the interactions I like the most, will share more of them soon, stay tuned.
aiforui.dev/lab
profile moved → @hipnopol retweeted
Damn. Someone rebuilt basically a whole paid AI video studio on ComfyUI, with ZERO API nodes. Open weights only, MIT licensed.
It's called Freefield. Built for Comfy's open source Higgsfield challenge: 117 apps covering 95 of the 129 tools and apps Higgsfield shows.
What the video shows:
• Cinema Studio: FLUX 3 Klein draws the look, LTX-2.5 moves it, dolly in, one take
• Extend Video: MiniMax H3 picks up where your clip ends
• A photo plus a face combined into one character
• Every app is just data: an app.json plus a ComfyUI workflow
• A lab that runs every app on a GPU, and a VRAM map showing what each app needs
The models under the hood: LTX-2.5, FLUX 3 Klein, MiniMax H3, Wan 2.1 and 2.2, Qwen-Image-Edit, Z-Image, SeedVR2, Stable Audio 3 and MiniMax Music 3.
Why it's sick for local rigs: the README lists which apps fit a 24GB card. Lots fit whole, like Create Image at 17.8GB, Edit and Inpaint at 11GB, and background removal at 0.4GB. LTX-2.5 video apps run tight at around 21.5-22.8GB. And any ComfyUI workflow can become a draft app with one command.
How to try it: clone neuregex/freefield, run python tools/local_backend.py setup, download the model packs you want, then npm run serve.
Honest caveat: some of the heaviest apps, like MiniMax H3 video with its 27.1GB text encoder, are marked as needing Modal's cloud GPUs. A 24GB card won't run everything.
I got featured by this new Instagram account promoting ai art and I tell you, I like the feeling I get when I read about my work, never happened before
36
Takiego sposobu na tworzenie muzyki z pomocą AI jeszcze nie widziałem. Nucę do telefonu, Claude, zapisuje nuty, wrzucamy je w syntezator (web audio -czysty kod, bez plików). Tak powstaje soundtrack do DITHER. Czy to nadal slop? Moim zdaniem nie, i nie zmienicie mojego zdania.
2
53


























