Preferred (Web-Safe) Profile
| Aspect | Required / Recommended Setting | Rationale |
|---|---|---|
| Container | MP4 (.mp4) | Universally supported in all browsers, mobile SDKs, and most broadcast playout systems. |
| Video Codec | H.264/AVC, Baseline, Main, or High profile, Level ≤ 4.2 (up to 1080p60) • CBR or 2-pass VBR • Average bitrate ≤ 10 Mb/s | Keeps files “analysis-ready” while preserving quality. Higher bitrates trigger server-side transcode. |
| Audio Codec | AAC-LC, 48 kHz, 2 channels, ≤ 320 kb/s CBR | Widely supported; maintains sync and loudness accuracy for speech models. |
| Pixel Format | 4:2:0, 8-bit YUV | Aligns with GPU decoding pipelines; avoids colour-space conversions. |
| Resolution | Any up to 4K (3840 × 2160). 1080p (1920 × 1080) recommended for fastest turnaround. | |
| Frame Rate | Constant frame rate (CFR) 23.976–60 fps. AI modules calibrate faster on CFR; VFR files are re-timed. | |
| Scan Type | Progressive; interlaced sources should be de-interlaced before upload. | |
| File Header | Fast-start / Web Optimised: moov atom placed at file head; correct duration metadata; indexed key-frames at GOP boundaries. | |
| GOP Structure | Closed GOP, ≤ 2 s (≈48–60 frames). Short, regular GOPs accelerate seek requests and temporal segmentation in analysis. |
Alternative Containers/Codecs Accepted (Transcoding Required)
| Container | Typical Codecs Seen | Action Taken |
|---|---|---|
| MOV | ProRes 422 / 444, DNxHD/R, Animation | Always transcoded to H.264/AAC. |
| MKV | H.265/HEVC, VP9, FLAC | Always transcoded to H.264/AAC. |
| WEBM | VP9 + Opus (bitrate > 10 Mb/s) | Re-encoded to compliant H.264/AAC. |
| AVI, MPEG-TS, M2TS, MXF, VOB, FLV, WMV, 3GP | MPEG-2, VC-1, WMV9, CineForm, others | Full transcode; longer queue times expected. |
Header & Stream Integrity Requirements
- Files must be free of DRM, encryption, truncation, or broken indexes.
- Time-code discontinuities (PTS/DTS jumps) should be corrected prior to upload.
- Closed captions or subtitle tracks are supported in MP4 (CEA-608/708 or WebVTT) and MKV/WEBM (SRT, WebVTT). Unsupported caption tracks are stripped during transcode.
Performance Impact Summary
| Upload Scenario | Playback Availability | AI Analysis Start | Notes |
|---|---|---|---|
| Preferred profile (MP4, H.264/AAC ≤ 10 Mb/s) | Instant | Within seconds | No server transcode. |
| Preferred container but bitrate > 10 Mb/s | Delayed (∝ file size) | After transcode | Bitrate down-conversion required. |
| Non-preferred codec/container | Delayed | After transcode | Full re-encode; queue length depends on resolution/codec complexity. |
Preparing Files
Recommended toolchain presets (example CLI flags shown for FFmpeg):Questions? For workflow-specific encoding recommendations (e.g., broadcast mezzanine archives, HDR pipelines, or high-frame-rate sports captures), please contact [email protected]. Our solutions team can provide tailored presets and guidance to balance quality, speed, and storage across your ingest pipeline.