Summary

This repository is a curated, visually-organized survey of Text-to-Video (T2V) generation technologies, covering commercial products, open-source models, research papers, datasets, and benchmarks. It aids engineers, researchers, and power users in navigating the rapidly evolving T2V landscape, facilitating informed decisions on adoption, forking, or building T2V solutions.

How It Works

The repository functions as a living catalog, meticulously organizing T2V advancements. It categorizes information into commercial platforms, generative models, AI avatar tools, creative suites, research papers (by year), and datasets/benchmarks. Key trends and comparative details are presented via tables and featured product cards, offering a structured, up-to-date reference for the T2V ecosystem.

Quick Start & Requirements

"Quick Start" sections provide git clone and pip install commands for open-source models like Wan 2.1 and HunyuanVideo 1.5, directing users to official guides for inference. Prerequisites are model-dependent, often requiring substantial GPU VRAM (8GB-24GB+), CUDA, and specific Python versions for local deployment.

Highlighted Details

• Key Trends (June 2026): Native 4K, synchronized audio, multi-shot storytelling, character consistency, and inference acceleration are standard. Open-source models now rival commercial systems.
• Featured Commercial Products: Includes Runway Gen-4.5 (professional, @reference consistency), Kling 3.0 (motion-heavy), Veo 3.1 (native 4K, broadcast), and Seedance 2.0 (multi-shot).
• Prominent Open-Source Models: Details Wan 2.7 (Apache 2.0, ~8GB VRAM), HunyuanVideo 1.5 (Apache 2.0, ~14GB VRAM), and LTX-2.3 (Apache 2.0*, ~8GB VRAM), including size, license, and VRAM.
• Extensive Datasets & Benchmarks: Curates resources like WebVid-10M, InternVid, and benchmarks such as VBench and T2V-CompBench for evaluation.

Maintenance & Community

Maintained by jianzhnie, the repository welcomes contributions via pull requests for new entries. Direct contact is available via the curator. No specific community channels (Discord/Slack) are listed.

Licensing & Compatibility

The repository itself is Apache 2.0 licensed, permissive for commercial use. However, individual listed models and products have their own varied licenses requiring separate consultation for specific compatibility and usage restrictions.

Limitations & Caveats

The T2V field evolves rapidly, with frequent feature/availability changes. Prominent consumer apps like OpenAI's Sora and Haiper were discontinued in early 2026. Many advanced open-source models require significant hardware resources (high VRAM GPUs), posing an adoption barrier.