Submit and view evaluations of video models
VLMEvalKit Eval Results in video understanding benchmark
Generate Talking avatars from Text-to-Speech
Dense Grounded Understanding of Images and Videos
Generate animated characters from images
Find frames in videos matching text queries
Generate summaries from YouTube videos or uploaded videos
Fastest high-quality video diffusion model.
input text, extracting key themes, emotions, entities,
Final Year Group Project : Video
Stream audio/video in realtime with webrtc
Generate videos from images and text prompts
Browse robotic datasets visually
The T2V-CompBench Leaderboard is a benchmarking platform designed for evaluating and comparing video generation models. It provides a centralized space for researchers and developers to submit their models, evaluate performance, and view results in a standardized manner. The leaderboard aims to promote transparency and progress in the field of video generation by enabling fair comparisons and fostering innovation.
• Model Submission: Users can submit their video generation models for evaluation.
• Standardized Evaluation: Models are evaluated using a set of well-defined metrics to ensure fairness and consistency.
• Real-Time Leaderboard: The leaderboard updates in real-time, showing the latest results and rankings.
• Detailed Performance Reports: Users receive comprehensive reports highlighting their model's strengths and weaknesses.
• Customizable Filters: The leaderboard can be filtered based on specific criteria, such as model type or evaluation metric.
• Open Access: The platform is designed to be accessible to both academia and industry, promoting collaboration and knowledge sharing.
What metrics are used to evaluate video generation models on the leaderboard?
The leaderboard uses a variety of standardized metrics, including but not limited to, video quality, temporal consistency, and diversity. These metrics are designed to provide a comprehensive assessment of model performance.
How do I submit my model for evaluation?
To submit your model, you must register on the platform, prepare your model in the required format (e.g., ONNX), and follow the submission guidelines provided on the website.
Is the leaderboard free to use?
Yes, the T2V-CompBench Leaderboard is free to use for academic and research purposes. However, users must register and agree to the platform's terms of use.