Video subtitles for streaming platform: one WebVTT file per language

When you open a video in the admin, the subtitle area sits alongside the other metadata for that title. You pick a language from the catalog, choose the WebVTT file from your computer, and save. The file is stored with the video and listed in its track set. Repeat for each language you have. Removing a track is the reverse: delete the entry and the player stops offering that language on the next load.

A series is a container; the episodes are the playable items, so subtitle uploads happen on the episode record rather than on the series. This is deliberate. Translation schedules rarely line up neatly, and a distributor often has French captions for the first six episodes before the last four are ready. Each episode carries its own track list, and an episode with no uploads simply shows no caption selector.

The admin holds a catalog of languages that the whole platform shares. Subtitle uploads are tagged from that list, and the label the viewer sees in the player comes from the same place. Keeping this list tidy is worth ten minutes of your time before launch: if two admins add "Portuguese" and "Português" as separate entries, viewers will see both, and one of them will usually be empty on any given title.

When a viewer starts a title, the player receives the video manifest and the list of subtitle tracks with their language labels. It shows a caption control, lists the languages, and lets the viewer choose one or switch captions off. Selection happens client-side; the video stream itself does not change. The player draws the cues at the times written in the WebVTT file, which is why timing must match the encoded video.

The upload form accepts a file and stores it; it does not parse or validate the contents. That means an SRT file with a .vtt extension will upload without complaint and then fail to render, with no error surfaced to the viewer. Treat the format rule as yours to enforce. A valid WebVTT file starts with the line WEBVTT, uses a period before milliseconds, and separates cues with blank lines. Most subtitle editors export it directly.

Flicknexs can run an AI transcription job on a video. The job returns the spoken text, the detected language, the duration and a list of timed segments, and the platform stores those as metadata used for search and summaries. It does not write a WebVTT file and it does not add a caption track to the player. If you want captions from that transcript, a person exports and times it in a subtitle tool, then uploads the result.

The AI transcription job produces metadata for search and summaries, not caption tracks. There is no button that turns a video into a WebVTT file, and there is no machine translation of an uploaded track into other languages. Human or third-party caption production remains part of your workflow, and the platform stores the result. Budget for it per title and per language.

The upload stores whatever you give it. SRT, TTML, SCC or a broadcast caption file will be accepted by the form and ignored by the player. Convert to WebVTT before upload and test playback after. Because nothing warns you, a wrong-format track can sit on a live title for weeks unless someone plays it with captions on.

Font, size, color, background and position of caption text are not exposed as operator settings. Captions render in the default style of each player, and viewers on some devices can adjust caption appearance from the device accessibility settings. If a brand guideline demands a specific caption look, that is a customization request, not a toggle.

The subtitle upload lives on video and episode records. Live streams and linear channels do not have a subtitle upload, and there is no live captioning path in the platform. If a live event needs captions, plan to caption the recording after the event and attach the tracks to the on-demand version.

The language catalog tags each upload with a single language, and a title holds one file per language entry. Variants such as "English for the hearing impaired" alongside plain English need separate catalog entries to coexist on one title. That works, but it means the catalog grows with every variant you decide to support, so decide the variant policy before the catalog fills up.