OCR embedded PGS and VobSub tracks into text subtitles #206
No Assignees
Notifications
Due Date
No due date set.
Dependencies
No dependencies set.
Reference: yolo/arr#206
Reference in New Issue
Block a user
Delete Branch "%!s()"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Deferred out of the Subtitles milestone deliberately; filed so the reasoning
survives.
PGS (BluRay) and VobSub (DVD) subtitle streams carry bitmap images of text, not
text. arr records them as satisfying their language for viewing — Jellyfin
renders them — but they can never feed a translator, so an English BluRay with
no external subtitle available has no translation source at all.
OCR would close that gap: extract the stream to
.sup, run tesseract or anequivalent over it, get an SRT that is both a real subtitle and a translation
source.
Cost of doing it: a new heavy binary dependency, per-image processing time that
is not spawn-and-die-cheap, and OCR errors that propagate silently into any
translation made from the result. DESIGN.md §15 states this as a non-goal, so
picking it up means amending §15 first.
#268 proposes a cheaper route to the same gap this issue describes, for the common case.
The premise here is "an English BluRay with no external subtitle available". On the file that surfaced this (a 2160p BluRay remux), an English subtitle was readily available from OpenSubtitles — arr never asked for it, because the embedded PGS track already marked English as satisfied. Fetching that text and aligning it to the disc's own packet timings gives a real translation source with disc-exact timing, no OCR and no new binary.
That does not close this issue, but it should shrink it: OCR remains the only option when no provider has a text subtitle in any language, which is where #207 also lives. Worth re-reading this one's scope once #268 lands.