CogVLM2-LLaMA3-Caption is an open-source multimodal model that analyzes video frames and generates accurate, context-aware textual captions or summaries in real time.
Ovis1.6 is an open-source multimodal large language model designed for joint text and image understanding, excelling in math reasoning, object recognition, and cross-modal analysis.
ActAnywhere is a research video generation model that automatically creates temporally consistent, motion-aware backgrounds for foreground subjects—enabling seamless background replacement and visual effects without manual keyframing or green screens.
FineVideo is a large-scale, open multimodal video dataset designed to advance AI models’ understanding of spatial-temporal scenes, emotions, narratives, and cross-modal relationships in video.
Vary-toy is a compact, open-source visual language model enabling document OCR, image description, visual question answering, and multimodal dialogue on consumer-grade hardware.
SafeEar is an open-source AI framework that detects audio deepfakes by analyzing acoustic features while preserving speaker privacy across multiple languages.
Llama 3.2 is Meta's open-weight multimodal model family optimized for edge devices, supporting vision-language understanding, on-device text generation, and customizable local deployment.
V-JEPA is an open-source, self-supervised video foundation model that learns visual representations by predicting masked frame features—enabling robust video understanding without labels.
HivisionIDPhotos is a lightweight, open-source AI tool that generates professional ID photos with automatic matting, background replacement, layout printing, and optional beautification — all running locally on CPU.
WiseFlow is an open-source AI workflow tool that automatically crawls, parses, classifies, and organizes web and social media content—including WeChat articles—using LLMs and statistical learning, all without GPU requirements.
F5-TTS is an open-source, high-performance text-to-speech tool that generates natural, expressive speech using flow matching and diffusion transformer models.
OmniParser is a lightweight, open-source screen parsing module that converts UI screenshots into structured, actionable elements to boost visual-language model performance in GUI-based agent tasks.
Seed-VC is a zero-shot voice cloning model that converts speech to new voices without needing target speaker recordings, hosted as a free Hugging Face Space.
CosyVoice 2.0 is an open-source, high-fidelity streaming text-to-speech model that delivers ultra-low-latency, multilingual, emotionally expressive, and speaker-controllable voice synthesis.
Fluid is a research-grade text-to-image autoregressive model that uses continuous tokens and randomized token ordering to generate high-fidelity, semantically aligned images.
Motionshop is an open-source framework that replaces real people in videos with expressive, lighting-aware 3D digital humans while preserving natural motion, timing, and visual continuity.