Help Center> VAS> API Reference> Before You Start> Constraints and Limitations on Using VAS
View PDF

Constraints and Limitations on Using VAS

VAS has certain restrictions, some cost-based and some technological. The system-wide restrictions affect all subservices, as well. In addition to system-wide restrictions, some subservices have additional, independent restrictions.

System-Grade Restrictions

  • Supported video formats: AVI, WMV, MPG, MPEG, MP4, MOV, M4V, MKV
  • Videos cannot be stored in OBS buckets encrypted by KMS.
  • The size of a single video cannot exceed 4 GB.
  • Supported frame rates (fps): 23.97, 24, 25, 29.97, 30, 50, and 60.
  • Supported GPU decoding video formats: H.264, H.265, MPEG2, MPEG4, VC1, VP8, VP9

    Encoding Format

    Maximum Resolution

    MPEG2

    1920 x 1080

    MPEG4

    1920 x 1080

    VC1

    2048 x 1024

    H.264

    1920 x 1080

    H.265

    1920 x 1080

    VP8

    1920 x 1080

    VP9

    1920 x 1080

Video OCR

  • When data is read from a specified URL, the video size cannot exceed 1 GB.
  • Numbers, English subtitles, and simplified and traditional Chinese characters can all be identified.
  • Horizontal and vertical text can be recognized, as well as many unclear or artistic fonts, but text arranged into a circle or viewed from a severe angle are typically not handled well.
  • The video resolution must be at least 300 x 300 pixels.
  • The video frame rate must be greater than 1 fps.
  • Supported regions: CN North-Beijing1 and CN North-Beijing4.

Video Celebrity Analysis

  • When data is read from a specified URL, the video size cannot exceed 1 GB.
  • The facial image must be at least 40 x 40 pixels.
  • Faces can be recognized with up to ±15° pitch and ±30° yaw.
  • A maximum of 20 faces can be identified at the same time.
  • Supported regions: CN North-Beijing1 and CN North-Beijing4.

Video Topics Segmentation

  • When data is read from a specified URL, the video size cannot exceed 1 GB.
  • This subservice only applies to standard news videos with news anchors identified.
  • A face can only be identified if the facial image includes at least 60 pixels.
  • Faces of news anchors must be in the valid area (excluding the 10% edge area).
  • A news anchor can be identified only when the front face of the news anchor appears for at least three times and retain on the image consistently for at least 2 seconds each time.
  • A maximum of 3 news anchors can be identified simultaneously.
  • Supported regions: CN North-Beijing1 and CN North-Beijing4.

Video Cover Selection

  • Video files in TS format are supported.
  • When data is read from a specified URL, the video size cannot exceed 1 GB.
  • GPU accelerated decoding supported for videos in H.264, H.265, MPEG4, VP8, and VP9 formats. The maximum resolution is 1920 x 1080 pixels.
  • Common frame rates are supported, such as 23.97, 24, 25, 29.97, 30, 50, and 60 fps.
  • Supported regions: CN North-Beijing1 and CN North-Beijing4.

Video Fingerprinting

  • Video files in TS format are supported.
  • When data is read from a specified URL, the video size cannot exceed 1 GB.
  • GPU accelerated decoding supported for videos in H.264, H.265, MPEG4, VP8, and VP9 formats. The maximum resolution is 1920 x 1080 pixels.
  • Common frame rates are supported, such as 23.97, 24, 25, 29.97, 30, 50, and 60 fps.
  • Supported regions: CN North-Beijing1 and CN North-Beijing4.

Video Content Moderation

  • The size of a single video cannot exceed 2 GB.
  • The video frame rate must be greater than or equal to 1 fps.
  • Currently, API calling concurrency cannot be ensured.
  • Supported regions: CN North-Beijing1, CN North-Beijing4, and CN East-Shanghai1.
Parent topic: Before You Start