Elysia Tools
Navigation
Workflow Playbook
Diagnose fixed offset, progressive drift or edit differences in existing translated subtitles, confirm cue correspondence, merge a bilingual track and verify readable selectable playback.
Hubs
This workflow is for a translator or editor with an existing reviewed translation and one specific target cut. It does not create a transcript or replace translation review. Inventory duration and existing subtitle streams, but also compare opening, middle and closing dialogue with the picture. Two files can have identical duration yet different edits. Keep originals and a small anchor table recording current cue time, correct spoken time and the observed error.
A constant error suggests a shared offset. An error that increases smoothly may support a linear correction; use two well-separated anchors and test a third independently. A jump after a removed scene calls for sectional repair or a correct-cut reference, not one global factor. A different frame-rate label does not by itself imply a faster or slower movie.
Choose one timing route first. The audio-based shifter requires matching audio even in manual mode; its automatic first-speech estimate is a heuristic, not a word alignment guarantee. Negative shifts can clamp early cues to zero, so inspect the opening again. For linear calibration, record factor and offset and check the middle as well as the fitted anchors. The retimer's SRT text includes a report preamble: keep that as evidence, but remove it from the clean subtitle deliverable through a checked editor export.
If a reference track already fits the exact cut, timing transplant can reuse its clock. Same cue counts permit ordered pairing, not semantic proof. Nearest-time matching is a candidate assignment; it can leave reference-language fallback text and unused translation cues. Read the report and inspect those passages. If scenes are missing or cues were split differently, repair the mapping manually rather than hiding the problem with a wider matching tolerance.
Set which language appears first and which clock is authoritative. Index merging belongs to tracks with confirmed equivalent segmentation; overlap-based pairing still needs dialogue review. Unmatched cues can remain single-language, so a valid output file may contain an incomplete bilingual experience. Preserve useful speaker identifiers and sound annotations when the audience needs them; do not remove accessibility information as cosmetic clutter.
Audit the merged text, not just each language separately. Both languages share screen space and reading time. Review long lines, dense exchanges and overlaps at the intended display size. Automatic overlap repair or duration extension can move captions across the next line or a scene cut; replay affected sections. Threshold profiles are review aids, not certification or proof of translation accuracy.
Export a clean SRT and keep source tracks, settings and exceptions beside it. Softmux produces one selectable bilingual track, not two independent language tracks, and copies video and audio only when codecs fit the chosen container. Existing subtitle streams may also be retained or converted; inspect the resulting stream inventory and labels instead of assuming the new track is first or correctly identified. Subtitle conversion may change styling, so compare the delivered display with the edited version.
Open the output in the intended player, turn the track on and off, seek to several anchors and inspect both language order and clipping. Record tested player, file revision and unresolved exceptions. This page covers timing repair and bilingual handoff; new translations and burn-in belong in the related subtitle workflow, and a full accessibility review belongs in the accessibility workflow.
Workflow playbook
Probe duration, streams and existing tracks, preserve source versions and record spoken anchors; metadata helps inventory the file but does not prove editorial cut identity.
Apply a measured manual shift using matching audio, a justified linear calibration or verified timing transplant; these are alternative routes, not three mandatory transformations.
Confirm language order and the authoritative clock, choose index matching only for equivalent cue segmentation, otherwise inspect overlap matches and every unmatched single-language cue.
Review reading speed, line lengths and overlaps, adjust with scene context, and export clean SRT; strip report preambles without deleting dialogue or accessibility information.
Compare several spoken anchors. Shift a constant error, calibrate a demonstrated linear drift, and repair cut-specific sections manually; different reported frame rates alone do not justify rescaling.
Transplant only from a trusted same-cut reference with verified dialogue correspondence. Equal cue counts are not proof; nearest-time matching cannot infer meaning or repair arbitrary splits.
The merger stacks languages in one track, not two selectable tracks. Use this route for an agreed bilingual display; separate languages or permanent burn-in need a different delivery path.
Add the reviewed SRT to a compatible MKV or MP4 while copying audio and video, inspect the resulting streams and labels, then manually test language order, seeking and sampled anchors.