This document defines the technical specifications and general requirements for all video and audio assets processed within the platform. The purpose is to ensure consistent input quality across all machine learning pipelines, enabling reliable analysis, feature extraction, and accurate model performance.
All submitted media files must adhere to the following guidelines regarding file size, structure, and encoding. These requirements are designed to minimize decoding issues, avoid data loss during processing, and maintain visual and acoustic integrity across different workflows.
The specifications are grouped into three main sections:
- General Requirements — Overall file structure and metadata prerequisites.
- Video Criteria — Technical standards for visual data such as resolution, bitrate, and codec.
- Audio Criteria — Audio encoding, sampling, and channel setup requirements.
Adhering to these criteria ensures that assets can be automatically validated, efficiently decoded, and seamlessly integrated into the machine learning and analytics pipelines without manual preprocessing.
General Requirements
Criterion |
Recommendation / Value |
Required |
|---|---|---|
| Filesize | Max. 10 GB per Video-File |
|
Duration |
≥10 Minutes |
|
| DRM / Encryption | No DRM Protection or other encryption methods. |
|
| File structure | One file per Asset containing audio and video data. Segmented Files are not supported. | |
No Watermarks / Logos |
Visual overlays may interfere with feature extraction models (e.g., object or brand detection). |
|
No Letterbox / Pillarbox |
Crop black bars to preserve only active image area. |
|
File Naming |
Use standardized format: |
|
Supplementary Data |
Provide poster, title, synopsis, genre, language, cast as separate data via API. |
|
Video Criteria
Criterion |
Recommendation / Value |
Required |
|---|---|---|
Resolution |
≥ 720p |
|
Bitrate |
Minimum 5 Mbps (higher preferred). Avoid excessive compression artifacts. |
|
Container Format |
Various Container Formats supported. |
|
Video Codec |
Various Audio and Video Codecs supported |
|
Framerate |
Constant frame rate (e.g. 24 or 25 fps). No variable frame rate (VFR). |
|
GOP Structure |
Short GOP preferred (e.g. 1–12 frames per keyframe) for frame-level accuracy. |
|
Duration |
Any, but should be >= 10 minutes only for Binge Markers product |
|
Color Space / Depth |
Rec. 709 / 8-bit (standard). Optionally Rec. 2020 or 10-bit for HDR workflows. |
|
Interlacing |
Always progressive (no interlaced content). |
|
Audio Criteria
Criterion |
Recommendation / Value |
Required |
|---|---|---|
Audio Codec |
Various Audio and Video Codecs supported. One File per Asset. |
|
Sampling Rate |
Minimum 44.1 kHz or 48 kHz. |
|
Number of Channels |
Mono or Stereo (2.0). Avoid multichannel (5.1 / 7.1) unless explicitly required. |
|
Loudness Normalization |
Target: -23 LUFS (for consistent speech/sound detection). |
|