Separate text decoding from font rendering. If accented letters or Chinese text turn into different characters when the subtitle file is opened, inspect its encoding and the receiver's interpretation. If the words decode correctly but show empty boxes on screen, inspect font support instead. Preserve the original text before converting a separate subtitle copy into the encoding required by the destination.
Inspect the earliest broken stage
Open the original subtitle in a capable text editor, then compare it with the imported caption text and rendered picture. Copy a short known phrase containing the affected characters for comparison. A character already replaced by a question mark may have lost information; changing fonts cannot reconstruct it reliably.
The Library of Congress SRT description notes the format's lack of a single character-encoding standard. UTF-8 is a useful explicit interchange choice when supported by the receiving application, but compatibility still needs an actual import test.
Test a small text sample
- Retain the original caption file and approved transcript.
- Identify whether decoding or rendering first changes the characters.
- Save a distinct copy in the receiver's documented encoding.
- Import a sample containing names, accents, Chinese and punctuation.
- Check both editable text and the final rendered captions.
Do not run repeated blind conversions on the same damaged file. Work from a clean original and keep each tested derivative identifiable. A preview containing only simple English letters can miss the fault completely.
Review timing after conversion too
Preserve sequence numbers, timestamp separators and blank lines while repairing text. An encoding fix should not become an accidental rewrite of the cue structure. Test a few cues from the beginning, middle and end after the receiving tool accepts the file.
Does renaming the extension fix encoding?
No. The extension does not convert the stored text. A renamed file can retain the same misinterpreted bytes.
What if the transcript is correct but the export is wrong?
Compare the export settings and receiving tool's import interpretation. Return to the correct transcript and regenerate a controlled sidecar copy rather than trying to infer every damaged name from the corrupted version.
Sources and scope
This is an original Imagild Editorial workflow guide, assisted by AI and reviewed against the linked sources on October 12, 2026. External tool procedures do not imply that Imagild provides those tools or guarantees their results.