AdSpark: A Large-Scale Dataset and Benchmark for Product-Centric Advertisement Video Generation
AdSpark releases a ~300K dataset and six-dimension benchmark for product-centric advertisement video generation.
AdSpark is a large-scale dataset and diagnostic benchmark for generating product-centric advertisement videos that keep product identity while telling multi-shot selling-point stories. AdSpark-300K contains roughly 300,000 reference image–prompt–video triplets, split into real-world and synthetic subsets, with structured annotations for identity, selling points, creative plans, and aligned audio. AdSpark-Bench evaluates six dimensions and finds current models struggle with product preservation, multi-shot storytelling, and selling-point visualization. Finetuning on AdSpark-300K improved representative models; the authors plan to release the dataset after acceptance.
- AdSpark-300K holds about 300K image-prompt-video triplets from real and synthetic ads.
- Annotations cover product identity, selling points, creative plans, and audio scripts.
- AdSpark-Bench scores visual quality, product fidelity, instruction adherence, temporal coherence, audio, and ad effectiveness.
- Finetuning on AdSpark-300K improved representative video models; the dataset will be released after acceptance.
- Products
- AdSparkAdSpark-300KAdSpark-Bench
From the original article
1 min read at huggingface.co
Product-centric advertisement video generation aims to create promotional videos that preserve fine-grained product identity while presenting selling points through coherent multi-shot narratives. However, this emerging task remains underexplored due to the lack of large-scale advertisement-specific datasets and comprehensive evaluation frameworks.
By Zhifei Yang, Zhao Jiang, Keyang Lu · huggingface.co
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.