AI

AdSpark: A Large-Scale Dataset and Benchmark for Product-Centric Advertisement Video Generation

Read full article at source ↗

Based on a feed excerpt; full article content has not been acquired. Source evidence retrieved 08/10/2026, 02:18:25 UTC.

Briefings

AdSpark is a new large-scale dataset and benchmark for product-centric advertisement video generation, derived from a major e-commerce platform. The dataset contains 300K reference image-prompt-video triplets, including real-world and synthetic samples with structured annotations for product identity, selling points, creative plans, and audio scripts. A diagnostic benchmark, AdSpark-Bench, evaluates generated ads across six dimensions: visual quality, product fidelity, instruction adherence, temporal coherence, audio alignment, and advertisement effectiveness. Experiments with AdSpark-300K-finetuned models confirm the dataset's effectiveness in addressing current research gaps.

Key points

• AdSpark-300K contains 300,000 reference image-prompt-video triplets from a major e-commerce platform.
• AdSpark-Bench evaluates generated advertisements across six dimensions including visual quality, product fidelity, instruction adherence, temporal coherence, audio alignment, and advertisement effectiveness.

Context

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. This emerging task remains underexplored due to the lack of large-scale advertisement-specific datasets and comprehensive evaluation frameworks. AdSpark addresses this gap by providing a unified dataset and benchmark for future research.

Limitations

The dataset relies on data from a single major e-commerce platform, limiting generalizability. The benchmark evaluates generated ads across six dimensions, but specific performance metrics for each dimension are not provided. The dataset was released upon acceptance, meaning the final release date is uncertain.

Generated by Bittrees-hosted qwen35-2b sha256:aaf42c8b7c3c on 04/11/2026, 08:07:48 UTC. Based on publisher-provided feed evidence; may not cover the full original article. Check the source for details.

View archived briefing on IPFS
bafybeignf63iaoindcp26suynrr2jpguaaogpcijuxskrkwt4ddcgkphqy