Open-source alternatives to StreamDiffusion

This project is an integrated software framework designed to facilitate generative image synthesis and high-performance model inference on Intel processor and graphics hardware. It provides a specialized inference engine that executes latent diffusion models to transform natural language descriptions into visual outputs. The library distinguishes itself by leveraging the OpenVINO toolkit to optimize machine learning models for specific Intel hardware architectures. By utilizing kernel-level hardware acceleration and static graph optimization, the framework improves execution throughput and re

This is a cross-platform media processing library that reads, writes, encodes, and decodes media in both browser and server environments. It supports common container formats including ISOBMFF, Matroska, Ogg, MPEG-TS, and HLS, and handles codec operations through a combination of WebCodecs API and WebAssembly-based encoders. Media is processed in streaming pipelines that maintain constant memory usage and automatically apply backpressure from output speed to all upstream components. The library distinguishes itself through a plugin-based codec registration system that allows extending support

GPAC is an open-source multimedia framework built around a pluggable filter graph pipeline, where modular processing units called filters connect into a directed graph to handle media workflows. At its core, the framework centers all media packaging and manipulation on the ISO Base Media File Format (ISOBMFF), with specialized tools for reading, writing, fragmenting, and encrypting MP4 and related containers. It also provides a declarative scene graph composition system for describing interactive multimedia scenes using MPEG-4 BIFS, X3D, SVG, or VRML syntax, alongside a hardware-accelerated re

This project is a scalable real-time communication platform designed for multi-party video and audio conferencing. It functions as a modular media server that orchestrates live sessions, manages virtual meeting rooms, and handles the distribution of media streams across distributed computing environments. The platform distinguishes itself through its support for protocol bridging, allowing for the translation between WebRTC, RTSP, SIP, and HLS. It incorporates hardware-accelerated transcoding to maintain performance across diverse devices and utilizes a plugin-based architecture to enable cus

stable-diffusion.cpp is a high-performance C++ inference engine designed for generating images and video from text prompts using Stable Diffusion models. It functions as a latent diffusion model runtime and a lightweight machine learning framework that enables local diffusion model execution on consumer hardware. The project distinguishes itself as a CPU-based image generator capable of running without a dedicated GPU. It employs a specialized C++ tensor backend and cross-backend hardware abstraction to dispatch compute tasks across different processor instruction sets and graphics APIs. The

TurboDiffusion is a video diffusion inference engine and generator designed to create high-resolution videos from text prompts and images. It provides a runtime environment for executing optimized diffusion model checkpoints with a focus on reducing latency and GPU memory usage. The project features a specialized training framework for aligning sparse-linear attention models with pretrained full-attention models. This system includes capabilities for sparse attention parameter merging and sparse-linear model alignment to reduce computational costs during inference while maintaining output qua

This project is a system-level utility for Linux that intercepts, modifies, and presents live webcam feeds as standard virtual video devices. By creating a bridge between physical hardware and user-space applications, it allows video conferencing software to consume processed streams as if they were native camera inputs. The software distinguishes itself through its ability to manage the lifecycle of video processing tasks as persistent background services. It monitors virtual device activity to dynamically allocate resources, ensuring that image processing and hardware usage are suspended wh

Nunchaku is a 4-bit model quantization library and diffusion model inference engine designed to run large-scale neural networks on consumer GPUs. It functions as a GPU-accelerated optimizer that reduces VRAM usage and increases inference speed through weight compression and memory management. The project utilizes low-rank weight decomposition and SVD weight quantization to compress models to four-bit precision while maintaining visual fidelity. It employs kernel-level operator fusion to minimize data movement and hardware-aware precision mapping to adjust numerical precision based on the unde

FramePack is a neural video synthesis engine and generation framework designed to produce long, temporally consistent video sequences. It functions as a diffusion model optimizer, providing a suite of techniques to manage the computational demands of high-parameter video models while maintaining visual stability during extended generation tasks. The system distinguishes itself through a hierarchical approach to frame prediction, which plans distant anchor frames before filling in intermediate content to prevent cumulative temporal drift. By utilizing constant-length context compression and to

React Native AsyncStorage is a persistent key-value storage library designed for React Native applications. It provides a unified local storage interface that works identically on both iOS and Android, ensuring saved data remains available across app restarts and when the device has no network connectivity. The library uses an asynchronous background I/O queue to handle all storage operations without blocking the JavaScript thread, communicating with native storage engines through React Native's bridge protocol. It includes a serialization layer that converts JavaScript values to strings for

XCGLogger is a logging framework for Swift applications designed to capture events and system state for debugging and troubleshooting. It automatically includes metadata such as timestamps, line numbers, function names, and filenames in every log entry. The framework minimizes CPU overhead through deferred string evaluation, which delays expensive interpolation until the active log level is verified. To prevent blocking the main execution thread during I/O tasks, it utilizes an asynchronous log router that offloads writing operations to background queues. The system supports multi-destinatio

FlexLLMGen is an inference engine and runtime designed to run large language models on a single GPU by combining weight compression with tensor offloading. It reduces model weight memory usage by approximately 70% through 4-bit quantization, and stores model parameters, attention cache, and hidden states across GPU, CPU, and disk to fit models larger than available GPU memory. The project distinguishes itself through a throughput-oriented batching approach that processes multiple generation requests together in large batches to maximize throughput on a single GPU. It also supports distributed

This project is a high-performance BERT embedding service and inference server designed to map text sequences into fixed-length numerical vectors. It functions as a machine learning microservice and distributed model server that decouples request handling from heavy computation. The system utilizes a ZeroMQ messaging infrastructure to provide low-latency communication between distributed clients and the inference server. It incorporates server-side batch processing and GPU workload scaling to maximize hardware utilization and manage high request volumes. The platform supports semantic search

LitServe is a Python AI inference server framework and LLM serving framework designed for high-concurrency inference. It functions as a distributed AI model server and dynamic batching inference engine, providing the tools to build and host custom servers that run AI models. The framework distinguishes itself through a dynamic-batching request queue that groups individual inference requests into single tensors to maximize GPU throughput. It supports distributed GPU scaling, allowing model workloads to be spread across multiple hardware accelerators to balance compute loads and increase total

LiveTalking is an interactive talking head engine and AI avatar management platform designed to synchronize synthetic speech with facial movements. It functions as a real-time orchestrator that connects large language models and text-to-speech services to neural-rendered digital humans. The project distinguishes itself through low-latency streaming capabilities and the ability to handle real-time conversational interruptions. It supports advanced audio-visual customization, including human voice cloning and the ability to drive avatar expressions using real-time webcam data. The platform cov

llm-d is a distributed serving framework designed for large language model inference. It functions as an inference orchestrator and gateway, providing a control plane for deploying model replicas and managing hardware accelerators. The system includes a batch inference scheduler and a cache manager to coordinate request flow and memory utilization. The project is distinguished by a disaggregated serving architecture that separates prefill and decode execution phases across specialized workers to maximize throughput. It employs a hardware-agnostic control plane and tiered cache offloading, mov

LightLLM is a high-performance serving framework for deploying and executing large language models. It functions as a multi-GPU inference engine and server capable of handling dense architectures, mixture-of-experts designs, and multimodal models that process both text and images. The system is distinguished by its specialized support for Mixture-of-Experts models using expert parallelism and fused kernels. It implements structured text generation through deterministic state machines and pushdown automata to enforce precise output formats. To optimize throughput, the framework employs specula

Sana is a framework for high-resolution image and video synthesis based on a linear diffusion transformer. It provides a toolkit for the training, fine-tuning, and execution of text-to-image and text-to-video models, as well as a video generative world model capable of simulating physical environments with precise spatial control. The project is distinguished by its use of linear complexity layers to handle high resolutions and its support for long-form, minute-length video generation in real time. It implements a two-stage inference paradigm that separates structural generation from visual t

This project is a framework for training and sampling generative models designed to produce high-quality images in few steps. It provides implementations for image generation models that transform random noise into structured visual data through an optimized sampling process. The system specializes in accelerating image generation through consistency distillation and consistency training. It includes tools to transform pre-trained diffusion models into faster versions by distilling knowledge from a teacher model into a student model, as well as methods to train consistency models from scratch

Lorax is a GPU-accelerated inference server and multi-adapter engine designed for serving large language models. It functions as a high-throughput system capable of deploying models via Kubernetes and managing the dynamic swapping of Low-Rank Adaptation adapters per request. The server distinguishes itself through multi-adapter dynamic batching, which allows requests using different adapter weights to be processed in a single GPU forward pass. It employs just-in-time adapter loading and weighted adapter merging to maximize throughput and enable multi-tasking without sacrificing performance.

This project is a PyTorch model serving framework designed to deploy and scale machine learning models in production via scalable network endpoints. It functions as a high-performance inference server, optimizer, and model lifecycle manager that handles model loading, request batching, and hardware acceleration. The system distinguishes itself through advanced orchestration and optimization capabilities, such as chaining multiple models into sequential workflows using execution graphs and employing dynamic batching to improve throughput and latency. It provides specialized support for generat

tiny-llm is a large language model inference engine and transformer model implementation. It serves as a quantized model runtime and paged key-value cache manager, providing a specialized inference stack optimized for Apple Silicon. The system distinguishes itself through high-throughput execution techniques, including continuous batching and paged attention. It utilizes a paged memory system to eliminate fragmentation during token generation and employs on-the-fly dequantization of compressed weights to reduce the memory footprint during matrix multiplication. The project covers a broad ran

exllamav2 is a high-performance inference library designed for running large language models locally on consumer-grade GPUs. It provides a GPU-accelerated runner and quantization tools to enable model execution without reliance on cloud-based computing services. The project features a quantization utility that compresses models into mixed bitrates between two and eight bits to reduce video RAM requirements. It distinguishes itself through a batched text generator that handles grouped requests and deduplicates cache data to increase throughput. The library covers a broad capability surface in

exllamav2 is a high-performance inference engine and framework for executing large language models locally on consumer-class GPUs. It provides a complete system for local model deployment, including a specialized inference engine and tools for model quantization. The project features a multi-GPU inference framework that distributes workloads across multiple graphics cards to run models that exceed the memory capacity of a single device. It includes a GPU model quantizer capable of converting models into mixed-precision formats between 2 and 8 bits to balance memory usage and accuracy. The en

This repository provides a collection of reference implementations and practical demonstrations for using WebRTC to establish real-time audio, video, and data communication. It contains code samples for negotiating peer-to-peer connections, managing media streams, and utilizing low-latency data channels. The project demonstrates the capture of audio and video from hardware devices, as well as the redirection of canvas element content into media streams. It includes examples of transferring arbitrary text and binary data between peers and managing the negotiation of direct connections. The sa

KServe is a Kubernetes-native platform for deploying and serving machine learning models as scalable inference services. It supports both generative AI models, including large language models, and traditional predictive models from frameworks such as TensorFlow, PyTorch, Scikit-Learn, XGBoost, and ONNX. The platform manages the full lifecycle of model deployments, including revision tracking, canary rollouts, A/B testing, and automatic rollbacks, and provides serverless scale-to-zero capabilities for cost-efficient resource management. KServe distinguishes itself through a standardized infere

This project is an MLOps architectural guide and framework for designing and deploying deep learning systems into production environments. It provides a structured approach to model inference deployment, ML pipeline orchestration, and the creation of production-level machine learning architectures. The project distinguishes itself through a focus on distributed deep learning and edge AI optimization. It covers methodologies for parallelizing model training across multiple GPUs to handle large datasets and applies techniques like quantization and distillation to reduce model size for embedded