franzipol

🚀 Major Update for ML-Sharp WebUI on Pinokio! 🌟 Native Mac Video Export (Zero CUDA) 🎥 10 New Camera Trajectories 💾 Low VRAM Mode (8GB GPU Support) ⚡️ Real-Time Previews Create stunning 3D gaussian splats from a single image. github.com/francescofugaz…

With all that often they happen It's very easy to finish the quotas on Antigravity: "Model produced a malformed edit that the agent was unable to apply" "AntiGravity Agent terminated due to error, You can prompt the model to try again or start a new conversation if the error persists"

I'm attempting to create a pinokio.co script and launcher for @openclaw which should make this much easier to install for us mere mortals. I'll check back in later with progress. Wish me luck. cc @cocktailpeanut

Qwen3-TTS Web UI on Mac! Qwen3-TTS web UI now works on Macs too, thanks to contribution from @PierrunoYT, who is also our Pinokio community mod. Thank you! Just install fresh and it should work, I've confirmed this working on my Macbook M1 Max here

Generate Songs with HeartMuLa, with 8GB VRAM! HeartMuLa is the latest open source song generation AI model everyone's raving about. But the default model couldn't run on low VRAM machines. Then WanGP just made it work with just 8GB VRAM! 1-Click Install on Pinokio.

Goodbye Creative Cloud subscription! We spent long time together until the price was raised up for AI credits that I've never used or I asked for. I won't miss you.

Introducing ShapeR, a method for robust conditional 3D shape generation from casually captured sequences. ShapeR leverages a rectified flow transformer conditioned on per-object multimodal data to turn casual image sequences into full metric scene reconstructions. Project Page: facebookresearch.github.io/ShapeR Paper: arxiv.org/abs/2601.11514 Links to code and huggingface below ⬇️

Used the Sharp Web UI (Single image to Gaussian Splat UI) by @franzipol running locally on my PC. Took the splat and took 2 snapshots while moving it around. x.com/cocktailpeanut…

FLUX.2 [klein] 4B + 9B are the newest BFL models in the Flux family—combining image generation + image editing in one compact architecture. Built for interactive workflows and quick iteration, with extremely fast inference on distilled variants. Two models. Two modes. ⚡️

@By0yom1 the glb is a mesh, it’s not used of 3dgs, 3dgs is like photogrammetry with steroids

@franzipol Yes, that's what I assumed too. I have tested one workflow with ComfyUI, which is creating .glb files from an image. But I didn't have a use case for that. And I didn't have blender installed. But I could install it 🤔. this was the image (AI creation by myself)

It took almost 2 years, but it was time for a proper update. 🕰️ 3DGS Converter v0.8 is here. The python CLI for Gaussian Splatting conversion. Now completely rewritten. ⚡️ GPU Accelerated (Taichi powered) 🔄 N-to-N Conversion: 3DGS, KSplat, SPZ, SOG, Splat, CC, Parquet, Compressed PLY. 🧹 Advanced Filtering (SOR, Density, Alpha) Any model. Any format. One command. 👇 Link in comments. #GaussianSplatting #OpenSource #ComputerVision #3DGS

@By0yom1 here is an example

@franzipol I am not familiar with these file formats. Do you use this with blender or similar graphic tools?

@By0yom1 just a python CLI utility for converting 3dgs files in various formats, nothing fancy but can be useful

@franzipol Sounds interesting! I have no idea, what it is. Tell me, please 😁.

i think the latest “cheap” prices I saw the other day was about 3100€ Pity that I’m broke, haha. My main setup is just a mini pc that I upgraded to 96gb ram ( ram kit that i bought for 230€ during summer) and a RTX 5060 TI bought at the same time for around 430 and it’s connected externally with Oculink. Even if not fancy GPU it was already a sacrifice to me

Sure 😅! I bought my new PC last year in August (the old one was 8 or 10 years old). And I wanted to create AI locally - so I bought this 5090, which was not that cheap (2800 Euro). But now it is probably extremely expensive. I wanted to run Linux with this brand new hardware: brand new GPU and also a brand new mother board. It took me 1 1/2 half weeks, but now I am very happy with this setup.

LTX-2 is amazing! WIth the distilled model on a RTX 5060 took less than 3 minutes for generating this 10 seconds 720p clip. I used Wan2GP on Pinokio

@By0yom1 Can I have that GPU too? 😂

@franzipol Great clip 😍! I got a RTX 5090 running with ComfyUI - OS LinuxMint. A 5 secs. clip is generated within 60 seconds with the normal version (not distilled). I had good results with the t2v workflow, but with the i2v results I am not that satisfied.

Swaps and enhances faces in images and videos github.com/allenk/facefus…

🚀 Major Update for ML-Sharp WebUI on Pinokio! 🌟 Native Mac Video Export (Zero CUDA) 🎥 10 New Camera Trajectories 💾 Low VRAM Mode (8GB GPU Support) ⚡️ Real-Time Previews Create stunning 3D gaussian splats from a single image. github.com/francescofugaz…

The manual annotation is now much more useful. The user experience is now like any other labeling platform: just draw a box around the subject to create a new bounding box. - Load your custom dataset - Label your images manually or automatically with DINOv3 - Optionally enrich your dataset - Train a YOLO model with your data - Test it immediately in the inference section - Export in ONNX or NCNN format for compatibility even on edge devices or smartphones > All for free. github.com/computer-visio…

Thank Ian 🔥

@Scannerian1 ❤️‍🔥

One that never got away! One of my buddies (@franzipol) put me on to this web UI on Pinokio for #SharpML. The code and instructions can be found here: github.com/francescofugaz…