Subtitle burn-in

Subtitle burn-in composites caption pixels into decoded video pictures before encoding an output whose subtitles are no longer separately removable.

Subtitle burn-in makes the subtitle's rendered pixels part of an output video. The pipeline decodes the relevant pictures, composes the subtitle presentation over them, and encodes the resulting pictures.

The original subtitle can remain selectable in the library. The burned-in output version no longer has a separate switch that removes those pixels.

One caption affects several pictures

Take a cue active from 2.25 to 3.25 seconds and video picture times n/24. Under the half-open interval model, pictures 54 through77 carry the caption: 24 pictures.

The caption image can be rendered or cached efficiently if it is static. Composition still has to include it in each applicable output picture. An animated caption can additionally change during that interval.

One rendered caption presentation is applied across the video pictures whose presentation times lie in its active interval.

At the end, the subtitle must disappear from subsequent pictures. Reusing the last nonempty overlay without respecting its lifetime can leave text stuck on screen.

Burn-in adds a video processing route

Stage Work
Decode Recover video pictures in a usable format/context
Subtitle rendering/decoding Produce current glyphs or graphics with required resources
Composition Combine subtitle pixels with each relevant picture
Encode Produce compatible output video
Delivery Supply that video and its remaining media components

A text decoder does not do all those steps. Neither does a hardware encoder eliminate the other stages.

For an Emby session, the selected route and requested presentation determine whether burn-in is used. Playback-method documentation is the source for product behavior; this page explains the work introduced by that choice.

Rendering and composition are distinct

A font lookup or shaping failure occurs before blending. A palette/alpha error affects the image supplied for composition. A coordinate transform can put otherwise correct lettering outside the retained picture.

Fixing the encoder setting does not resolve those subtitle-input relationships. Compare the current rendered layer, placement and timing before the final encode.

Colour treatment also matters for HDR/SDR. The subtitle's intended display colour must be interpreted in the selected working/output model. Overlay placement relative to tone mapping can affect its appearance; the pipeline needs a deliberate ordering.

Hardware can accelerate parts of the work

Hardware decoding, GPU composition and hardware encoding describe different stages. A CPU-rendered subtitle image can be uploaded and composed with device-resident video. Another route can download full pictures for CPU composition and upload them again.

Pipeline fundamentals develops those memory and timing relationships. Burn-in describes the resulting permanent picture content, rather than one fixed hardware backend.

See Also

References