Data/LLM & Multimodal Data

Multimodal Data at Unprecedented Scale

Explore our data catalog spanning audio, image and video. All data is ethically sourced, quality-validated, and ready for foundation model training.

Audio & Speech Data

Hundreds of millions of hours covering all languages and scenarios.

Data TypeDescriptionScale
English Raw AudioPodcasts, conversations, interviews, radio dramas, novels, film/TV, anime, etc.300M+ hours
Minority Language Raw AudioSpanish, Arabic, Portuguese, French, German, European languages, Japanese, Korean, Southeast Asian languages, etc.200M+ hours
Dialect Raw AudioRegional accents and dialects across multiple languages and regions, etc.100M+ hours
English TTS (Filtered)24k+ sample rate, 96–98% accuracy, podcasts/conversations/film & TV60M+ hours
English 4o (Filtered)Includes speaker info and timestamps, high-quality subsets60M+ hours
Music & SongsLossless music with lyrics and timing info, 128/256kbps+200M+ tracks
Sound EffectsTagged or described audio, wav/mp3 formats300M+ items

Image & Vision Data

Tens of billions of images covering all categories and scenarios across the web.

Data TypeDescriptionScale
Web-Scale ImagesPlatform-sourced imagery across categories, without descriptions20B+ images
Image-Text DatasetsOpen-source datasets (COYO, DataComp, OBELICS, etc.), image + text descriptions, interleaved20B+ pairs
Fine-Grained CategoriesAnimals, plants, cars, everyday objects, covering 10,000+ keywordsBillions
Premium Image Library2048×2048+ resolution with English labels and high-quality descriptions100M+ images
Real Portrait Data1080p+, single person, real photography100M+ images
Advertising Images1024+ resolution, with descriptions, ad copy, and creative info100M+ images
App Screenshot DataMobile/desktop software page screenshots across categories20M+ images

Video & Motion Data

Billions of video clips covering film, stock footage, digital humans, and all content types.

Data TypeDescriptionScale
Stock Video FootageProfessional stock and platform-sourced footage, 720p+Billions of clips
HD Video Footage1080p+, 6–60s, no subtitles/watermarks, nature/sports/scenery1B+ clips
Film & TV / Short DramaMovies, TV series, speeches, training, interviews, documentaries40M+ hours
Digital Human (Raw)Streamers, YouTube/TikTok, multi-language, 720p+100M+ hours
Digital Human (Clips)Single person on screen, no scene cuts, 3–30s segments, 720p+60M+ clips
Advertising Video1080p+, with descriptions, ad copy, and creative info50M+ clips
Open-Source DatasetsPexels, WebVid-10M, Koala-36M, Panda-70M, etc.Multiple major sets

Proprietary Data Platform

about.collection.title

    about.annotationPlatform.title

      about.delivery.title

        50B+

        Data Assets

        30+

        Languages

        30+

        AI Enterprise Clients