Someone just built a while film studio on top of ComfyUI.
IV studio is an open source Windows app that drives your ComfyUI and MiniMax H3. A 6 step wizard goes from idea to script, cast, storyboard, animatic and finished short film. Amazing! ⬇️
A free open source AI agent that lives inside ComfyUI! 🔥
AgentY runs on any LLM, even fully local with Ollama. It build workflows from what it finds on the web, then judges its own outputs against your brief and re-renders what it misses.
1
1
138
Sooo smooth! A toy Ferrari on a train windowsill becomes a real car on the road in one transition.
A creator made it with Minimax H3 and an RTX 5070Ti, at 1080p in about 8-9 minutes
118
GIMP just got Photoshop's AI tricks.
OpenFill is a beta GIMP 3 plugin that uses ComfyUI for generative fill, remove and expand.
No node graph needed and nothing leaves your machine ⬇️
96
These character swaps are getting ridiculous! 😂
A new MiniMax H3 LORA replaces the main character in a video and keeps the motion, expression, camera and scene
1
206
Imagine a Zoom call with Nikola Tesla, and he's running at 90+ FPS.
Someone got a realistic talking avatar running locally, in REAL TIME, on one RTX 4090. No cloud needed.
77
A random 10 seconds AI scene got over 200k views, so its creator is now racing to turn it into a short film before Halloween.
4 mercs in a post-human world, occult + cyberpunk. It's looking good!
OG authors in the end credits
81
Your handwriting as a font, made by Claude. Someone gave Opus 5.5 120 handwritten pages and turned them into a typeface ⬇️
(no comments on the handwriting itself)
103
Want a faster MiniMax H3 character swap? Someone timed the speed-ups on an RTX 5090, 15s of video at 1MP.
Turbo LORA alone, 632s.
Turbo + VEDA 353s, almost half.
205
Is this the FIRST diffusion LLM ever? Not quite, it's inception's fourth Mercury.
But Mercury 3 writes whole answers in parallel, not token by token, at 1068 tokens/sec. On it's own tests it leads 8 out of 12 benchmarks.
88
Soooooo now we have NVIDIA video super resolution on ComfyUI??? ⬇️
Built for long videos, it works from disk to keep RAM low. AI upscaling, denoising, deblurring, NVENC H.264/HEVC encoding, keeps the audio and with a built-in preview.
94
Fizgig 8 looks amazing! And it finally has a web UI.
A free, open-source app for training LoRAs and full fine tunes for Klein 9B, Krea 2, Qwen Image 2.1, SDXL and more.
No more juggling machines, one app runs your PC, your other PCs and RunPod.
1
90
Scumble looks like a proper tool for FLUX inpainting. An open source desktop editor, draw a box, say what goes in it, and FLUX fills it as its own layer.
Locally on Comfy or remotely with Black Forest Lab key.
78
2026, the year AI and traditional VFX became one.
In this Evangelion fan test, an AI Sachiel is pinned under a CG Eva unit through a complex camera move with contact points locked on every frame. Amazing.
OG authors in the end credits.
1
123
Soooo, is Google back in the game? Gemini 4 Argon is out at $2/M input and $10/M output, half the price of Opus 5.5
BUT! It still trails Opus 5.5 and Muse Wonder and the pricesl doubles in December 🫠
2
1
155
Anthropic is putting 5% of its shares into a new Public Benefit Fund, and its co-founders have pledged to donate 80% of their equity.
Genuine move or smart PR?
91
DeepSeek just dropped V4.1 Flash 1017 as open weights, and it jumps from 31.2 to 42.3 on Terminal-Bench 4.0, ahead of GPT-5.6 Sol.
Shorter outputs, faster task runs, and once again a big cut in cost per task.
1
72







