The system latency in Deep Live Hub can vary significantly depending on the configuration of your live stream workflow. It can range from sub-second delays (for transcripts in the input language using HLS real-time in partials mode) to up to 60 seconds when using the Live Editor with burned-in subtitles.
The graphic below illustrates the different delays based on the chosen workflow configuration, whether you are streaming video or text only.
Delays Based on Workflow Configuration:
-
V1: Live Stream with Video Output
- Transcription: Takes between 9-25 seconds to generate, depending on the infrastructure.
- Live Editor (optional): Adds approx. 60 seconds if activated for human review.
- Translation (optional): Adds another +5 seconds for generating translated subtitles.
- Post-Editor (optional): Adds an additional +60 seconds for a final review of subtitles.
- Output Delay: Adds up to 12 seconds if using RTMP, RTMPS, or HLS with burned-in subtitles or closed captions, also depending on the infrastructure.
-
V2: Broadcast with Text-Only Output
- Transcription: Takes between 5-25 seconds to generate, depending on the infrastructure.
-
- Live Editor (optional): Adds approx. 60 seconds if activated for human review.
- Translation (optional): Adds +5 seconds for translated text output.
- Output: Text-only output formats such as HTTPS, WebVTT, or EBU-TT.