Image · Video · Audio

An open-source system that turns AI coding assistants into autonomous video studios, handling research, scripting, asset generation, and final render.

A minimal C/C++ port of OpenAI’s Whisper built to transcribe speech locally on phones, browsers, and underclocked POWER9 boxes.

VibeVoice is a family of open-source speech models from Microsoft built to ingest, transcribe, and generate very long audio sessions—up to an hour—in a single pass.

An offline, cross-platform dictation app that aims to be the most forkable speech-to-text tool, not the most polished one.

A Tencent research project that restores degraded faces by tapping into the rich priors locked inside a pretrained StyleGAN2 model.

RVC exists because voice conversion usually demands heavy compute and hours of data; it packs training, inference, and vocal separation into a browser UI that claims to work with ten minutes of audio and weaker GPUs.

Deep-Live-Cam exists to turn a single selfie into a real-time webcam deepfake or video face swap running entirely on local hardware.

It wraps Stable Diffusion—and the whole messy ecosystem of models, upscalers, and LoRAs—into one browser-based workbench.

LocalAI wraps 36+ inference engines behind one OpenAI-compatible API and pulls them on demand, so you can run LLMs, vision, voice, and video on anything from a CPU to a Jetson.

It renders high-quality novel views of real-world scenes at 30 fps by replacing costly neural radiance fields with optimized 3D Gaussians.

It collects the best social-media experiments with a Gemini-2.5-flash-image derivative and releases a 150k identity-consistent dataset for the community.

SGLang exists to push low-latency, high-throughput inference for LLMs and multimodal models from a single GPU up to massive clusters.

Because you shouldn't need to reverse-engineer a black box just to swap a noise scheduler or fine-tune a diffusion model.

A reimplementation of OpenAI's Whisper that trades the original inference engine for CTranslate2 and gains up to 4× speed without sacrificing accuracy.

Trained its own deep-learning models so you don't have to sing along.

It turns face swaps and lip-syncs into queued, retryable batch jobs instead of one-off scripts.

It gives artists and professionals a local, node-based studio for Stable Diffusion, treating AI as a collaborator rather than an opaque generator.

OpenAI's Whisper is accurate but slow and timestamp-imprecise; WhisperX bolts on batching, forced phoneme alignment, and speaker diarization to fix that.

CLIP learns shared image-text representations so you can label photos with natural language instead of curated datasets.

CosyVoice provides open-source training, inference, and deployment tools for zero-shot multilingual speech synthesis using large language models.