Week of Oct 5 to Oct 11, 2026
Tools of the Week
Every Monday this page reshuffles and puts a different tool forward from each of the 27 categories in the directory. It is a way to surface the good work that never reaches the top of a ratings list. 27 picks are live right now, and the next rotation lands Monday, October 12.
Inria's reference code for 3D Gaussian Splatting, turning posed photo sets into real-time scenes.
Microsoft SDK for building LLM agents and plugin-driven workflows in C#, Python, and Java.
Audio-driven Tencent model that animates avatar images into emotion-controllable dialogue videos.
AI coding assistant that works with your existing codebase from the terminal.
Collection of Kubernetes-native projects for pipelines, training, tuning, notebooks and model management.
High-level deep learning API that runs the same model code on JAX, TensorFlow or PyTorch.
Interactive image editing system driven by brush strokes with AI intent prediction.
Meta's layered guardrail framework that scans agent prompts, plans, and generated code.
Batched Whisper transcription with phoneme-based word alignment and optional speaker labels.
Contrastive language-image pre-training model by OpenAI for zero-shot visual classification.
Computer vision annotation tool by Intel for image and video labeling.
Desktop workbench for synthetic data generation, evals, fine-tuning, and RAG with Git based collaboration.
ComfyUI node pack that detects faces and regions, then re-renders them for sharper detail.
Pixel-space diffusion model by DeepFloyd/Stability AI with strong text rendering.
MIT-licensed 671B mixture-of-experts reasoning model from DeepSeek, plus distilled versions from 1.5B to 70B.
Platform for serving LLM, embedding, speech, and image models behind one OpenAI-compatible API.
GPT-2 training implemented in plain C and CUDA with no PyTorch dependency, by Andrej Karpathy.
Stability AI training and inference code for generative audio models including diffusion and LMs.
Topic modeling library leveraging transformer embeddings and c-TF-IDF.
OCR model that converts PDFs and images into HTML, Markdown, or JSON with layout, tables, and handwriting.
All-in-one desktop and Docker app for private LLM chat with your documents.
IBM's speech-aware Granite LLMs for transcription and speech translation in six languages.
Text-to-speech system built on top of Whisper encoder representations.
Local desktop and browser app for running GGUF and other open models, with an OpenAI-style API.
High-performance embedding server by Hugging Face for production deployment.
NVIDIA's open world foundation models that generate physics-aware video for robotics and autonomy.
Coqui's multilingual voice cloning model, abandoned upstream, kept running by the Idiap fork; non-commercial only.