Shurshan/vintage-photos-for-video-generation

Vintage & Historical Photos for AI Image-to-Video Generation

Dataset Overview

This dataset contains a diverse collection of high-resolution, public-domain historical photographs, including street scenes, landscapes, and group portraits. It is specifically curated to help researchers and developers test and benchmark Image-to-Video (I2V) generative models (such as Stable Video Diffusion, Gen-2, Sora, etc.) on challenging vintage textures like film grain, sepia tones, and low contrast.

Primary AI Use Cases

• Image-to-Video Generation: Testing cinematic camera motion and scene animation on static historical photos.
• Temporal Consistency: Evaluating how well video models handle film scratches and vintage noise over time.
• AI Scene Animation: Bringing historical moments, moving elements (water, smoke, vehicles), and people back to life.

Generate Video from Old Photos Instantly

This open-source dataset is curated and maintained by Animate Old Photos.

If you are a creator, historian, or general user who wants to instantly turn static old photos into dynamic, high-quality videos without writing any code or setting up complex AI pipelines, try our dedicated online platform: Turn Your Old Photos into Videos Here

File Structure

• Contains 20+ historical scene and landscape images.
• Format: JPG/PNG
• Optimized for standard I2V aspect ratios and resolutions.

License

Images are sourced from the public domain. The dataset is free for academic research, AI model training, and non-commercial testing.