Work with transcriptsFree · No account needed
Subtitle vocabulary builder with word counts and context
See which words recur in a transcript you want to study. Start with frequency and the original context, remove words you already know, and add your own meanings before exporting a personal vocabulary list.
Input Subtitle or plain text · Optional known-word list · Your own definition notesOutput Word frequency and example contexts · First supplied timestamp per word · CSV and TSV vocabulary lists
Your subtitle text
Make the list yours
Words are lowercased and counted across all cues. Known words are excluded with the same word-splitting rules. Tokens longer than 200 characters are excluded with a warning. No definitions, translations, or proficiency levels are generated.
Choose a file up to 5 MiB and 2 million characters, with at most 10,000 cues or lines.
Words in context
| Word | Count | Example context | Your definition or note |
|---|---|---|---|
| Your vocabulary list will appear here. | |||
The list shows 50 words per page and up to three distinct contexts per word. Exports include all words, contexts, first available timestamps, and your notes. Everything is held in memory until you leave or clear the page.
How to use it
- Paste a transcript or choose a local SRT, VTT, or TXT file. Subtitle timecodes are parsed separately from the words, and recognized caption formatting is removed.
- Add an optional list of known words and set the minimum word length. Build the list to count the remaining words, ordered by frequency.
- Review up to three contexts per word and add a definition or study note in your own words. Browse the list in pages of 50, then download all rows as CSV or TSV.
A small example
The sample transcript uses “question” three times and “answer” twice. Adding “question” to your known-word list removes it on the next build. A note such as “a response that you can test” can be attached to “answer” and travels with its contexts in the export.
What to know
- This is a word-frequency and context tool. It does not translate, infer meanings, rank difficulty, assign CEFR levels, or decide whether a word is useful for you.
- Counting uses the browser’s Unicode word segmentation when available. Case and compatible Unicode forms are normalized; inflected forms remain separate words. A simpler Unicode splitter is used in older browsers, with an explicit warning.
- Known-word matching is exact after normalization. Multiword expressions are split into words. Digit-only tokens are excluded. Unbroken tokens longer than 200 characters are excluded with a warning, while spoken non-speech labels such as Music may remain if present in the transcript.
- Maximum: 5 MiB, 2 million transcript characters, 10,000 cues or lines, 10,000 distinct retained words, and 200,000 known-list characters. Manual notes allow 500 characters per word.
- Each word keeps at most three distinct short contexts and its first available subtitle timestamp. Plain text has no inferred timing. CSV/TSV downloads include all rows; notes are not saved after leaving or clearing this page.
References & format documentation
For assistants and developers
This is a browser tool with visible, labeled controls and local processing. Read the usage guide or machine-readable catalog for supported inputs, outputs, examples, and limits. No API key or server endpoint is available. A catalog listing is documentation, not automatic integration with every AI assistant.