Text-to-Video¶
Video
Open Source¶
-
Cosmos 3: Omnimodal World Models for Physical AI Discovered by AI
2026
An open-source omnimodal world foundation model family designed to jointly process and generate language, images, video, audio, and action sequences. It features a mixture-of-transformers architecture optimized for physical AI, world simulation, and high-fidelity video generation.
multimodalopen-weightsvideo-generationworld-model -
LTX-2 Discovered by AI
2026
An open-source foundational joint audio-visual model capable of generating high-quality, temporally synchronized audiovisual content. It features an asymmetric dual-stream transformer with a 14B-parameter video stream and a 5B-parameter audio stream.
audio-visualdiffusionditmultimodalGitHub · arXiv · Hugging Face
-
HunyuanVideo 1.5 Discovered by AI
2025
An 8.3-billion parameter open-source video generation model featuring a Diffusion Transformer (DiT) architecture with selective and sliding window attention, designed to run efficiently on consumer-grade GPUs while maintaining high visual quality and motion coherence.
ditopen-sourcetext-to-videovideo-generationGitHub · arXiv · Hugging Face
-
HunyuanVideo-1.5 Discovered by AI
2025
An upgraded, lightweight iteration of Tencent's video generation model, optimizing inference speed and resource consumption while maintaining high visual quality and motion dynamics.
lightweighttext-to-videovideo-generation -
LTX-2 / LTX-2.5 Discovered by AI
2025
An open-weights video and audio generation model family by Lightricks. It features native multi-shot scene generation, synchronized audio-video synthesis, and advanced prompt adherence using a Gemma-based text encoder.
audio-videoopen-weightstext-to-videotransformer -
NVIDIA Cosmos Discovered by AI
2025
An open platform of world foundation models, datasets, and tools designed for Physical AI and video generation. It provides highly efficient video tokenizers and diffusion-based world models that excel in generating physically consistent videos.
diffusionphysical-aivideo-generationworld-model -
NVIDIA Cosmos Discovered by AI
2025
An open platform of world foundation models designed for physical AI and world simulation, supporting high-fidelity text-to-video, image-to-video, and video-to-video generation.
open-weightsphysical-aivideo-generationworld-model -
Wan2.1 Discovered by AI
2025
An advanced open-source video foundation model suite supporting high-quality text-to-video and image-to-video generation up to 1080p resolution. It features a novel 3D variational autoencoder (VAE) and a scalable diffusion transformer architecture.
diffusion-transformeropen-sourcetext-to-videovideo-generation -
Wan2.2 Discovered by AI
2025
An advanced open-source video generative model featuring a Mixture-of-Experts (MoE) architecture that scales model capacity while maintaining low computational costs.
moetext-to-videovideo-generation -
Allegro Discovered by AI
2024
An open-source text-to-video generation model that produces high-quality, temporally consistent 6-second videos at 15 FPS and 720p resolution.
diffusiontext-to-videovideo-generation -
CogVideoX Discovered by AI
2024
A family of open-source text-to-video diffusion models utilizing an Expert Transformer architecture to align text and video representations. It generates coherent, high-frame-rate videos with significant motion and strong temporal consistency.
diffusion-transformeropen-sourcetext-to-videovideo-generation -
HunyuanVideo Discovered by AI
2024
A large-scale open-source video foundation model featuring a 13-billion parameter diffusion transformer architecture designed to compete with closed-source video generators. It supports high-fidelity motion, complex prompt adherence, and native 3D causal attention.
diffusion-transformeropen-sourcetext-to-videovideo-generation -
HunyuanVideo Discovered by AI
2024
Tencent's large-scale open-source video generation framework built on a 3D Causal VAE and a dual-stream diffusion transformer architecture, capable of producing high-resolution, temporally consistent videos from text prompts.
diffusion-transformertext-to-videovideo-generation -
LTX-Video Discovered by AI
2024
A highly efficient, real-time video generation model based on a Diffusion Transformer (DiT) architecture with only 2 billion parameters. It is capable of generating high-quality 30 FPS videos faster than real-time playback.
diffusion-transformerreal-timetext-to-videovideo-generation -
Lumina-T2X Discovered by AI
2024
A unified framework based on Flow-based Large Diffusion Transformers (Flag-DiT) designed to transform text instructions into images, videos, multi-view 3D objects, and audio clips.
diffusion-transformertext-to-videovideo-generation -
Mochi 1 Discovered by AI
2024
An open-source 10-billion parameter video generation model built on an Asymmetric Diffusion Transformer (AsymmDiT) architecture. It delivers high-fidelity motion, strong prompt adherence, and is released under a permissive Apache 2.0 license.
diffusion-transformeropen-sourcetext-to-videovideo-generation -
NVIDIA Cosmos Discovered by AI
2024
An open platform of world models, datasets, and tools designed for physical AI, robotics, and autonomous systems. It includes state-of-the-art video tokenizers and diffusion models trained on over 20 million hours of video.
open-sourcephysical-aivideo-generationworld-model -
Open-Sora Discovered by AI
2024
An open-source initiative by Colossal-AI to democratize efficient video production, reproducing the architecture of advanced video generation models like OpenAI's Sora. It supports scalable training and high-quality video synthesis.
diffusion-transformeropen-sourcetext-to-videovideo-generation -
Open-Sora-Plan Discovered by AI
2024
An open-source project aiming to reproduce OpenAI's Sora video generation model, utilizing a 3D Causal VAE and a diffusion transformer architecture to generate high-quality, temporally consistent videos.
diffusion-transformeropen-sourcetext-to-videovideo-generation -
Pyramid Flow Discovered by AI
2024
An open-source video generation model based on pyramidal flow matching, which enables efficient training and inference of high-resolution videos by progressively downsampling and upsampling latent representations.
flow-matchingopen-sourcetext-to-videovideo-generation -
Vidu Discovered by AI
2024
A high-performance text-to-video generator based on a diffusion model with a U-ViT backbone, capable of producing highly consistent and dynamic 1080p videos.
diffusionhigh-consistencyt2vu-vit -
2023
Adds a plug-in motion module to personalized text-to-image diffusion models, animating existing image checkpoints and LoRAs into short clips.
diffusionmotion-module
Proprietary / Industry Tools¶
-
Runway Gen-4.5 Discovered by AI
2026
Runway's next-generation proprietary video foundation model, setting new standards for dynamic action generation, temporal consistency, and precise controllability across diverse generation modes.
commercialproprietarytext-to-videovideo-generation -
Sora 2 Discovered by AI
2025
OpenAI's advanced video generation model that features improved physical accuracy, realistic motion, synchronized dialogue, and sound effects.
audio-synctext-to-videovideo-generation -
Google Veo Discovered by AI
2024
Google's state-of-the-art proprietary video generation model capable of producing high-fidelity, cinematic-quality videos with realistic physics, fluid motion, and natively generated audio.
proprietarytext-to-videovideo-generation -
Hailuo AI Discovered by AI
2024
A state-of-the-art proprietary video generation platform powered by MiniMax's advanced video models, capable of producing high-definition cinematic clips with smooth motion.
proprietarytext-to-videovideo-generation -
Kling AI Discovered by AI
2024
A state-of-the-art commercial AI video generation platform developed by Kuaishou Technology. It supports native 4K video generation at 60fps, multi-shot storyboarding, precise motion control, and synchronized audio generation.
4k-videocommercial-tooltext-to-videovideo-generation -
Luma Dream Machine Discovered by AI
2024
A commercial next-generation video generation model by Luma AI that creates high-quality, realistic cinematic shots from text instructions and images with fast generation speeds and consistent physics.
cinematiccommercial-tooltext-to-videovideo-generation -
MiniMax AI Video Generator Discovered by AI
2024
A powerful proprietary video generation model developed by MiniMax (also known as Hailuo AI) that excels at producing high-resolution videos with exceptional clarity and realistic motion.
proprietarytext-to-videovideo-generation -
OpenAI Sora Discovered by AI
2024
A state-of-the-art text-to-video model developed by OpenAI that can generate high-quality videos up to a minute long while maintaining visual quality, realistic motion, and strong adherence to complex textual prompts.
proprietarytext-to-videovideo-generation -
Sora Discovered by AI
2024
OpenAI's advanced video generation platform that simulates the physical world in motion, generating highly realistic and physically consistent videos from text and image prompts.
proprietarytext-to-videovideo-generation -
Vidu Discovered by AI
2024
A professional AI video generation platform developed by ShengShu Technology that supports text-to-video and image-to-video generation with high visual consistency and smooth motion.
proprietarytext-to-videovideo-generation -
Vidu AI Discovered by AI
2024
A high-performance AI video generator capable of producing realistic, high-definition videos from text and image prompts with consistent character movements and physics.
image-to-videotext-to-videovideo-generation -
A commercial video generation tool for text- and image-driven short clips, including anime-aware styles.
commercialvideo-generation -
A commercial text/image-to-video generation and editing suite used for previs, motion design, and stylized video in production.
apicommercialvideo-generation
