ElevenLabs|Scribe v2
Explain transcription through concrete text operations. Give each capability a clear input and change: time location, speakers, proper-term correction, and entity classification.
View HTML source
<!doctype html><html lang="zh-CN"><meta charset="utf-8"><meta name="viewport" content="width=device-width,initial-scale=1"><title>ElevenLabs|Scribe v2 · sequence</title><style>html,body{margin:0;width:100%;height:100%;overflow:hidden;background:#f7f6f0}canvas{display:block;width:100%;height:100%;object-fit:contain}</style><body data-mode="sequence"><canvas width="1000" height="600"></canvas><script src="data.js"></script><script src="common.js"></script><script src="scene.js"></script><script src="bridge.js"></script></body></html>Motion and composition
Local motion · Composition rules
Use versions and metrics only to establish context
Open with a regular grid, tiles, and a version mark to establish product identity, then show comparable metrics. Keep chart baselines, units, and labels consistent; bar heights do not replace necessary explanations.
Use mainly black type on white, with dark sections changing the rhythm. Warm colors or green mark current words and annotations; use sans serif suited to extended reading, keeping numbers and timecodes aligned.
Unfold text along time
Divide the transcript into speaker paragraphs, with a fixed timeline linked to the current paragraph's time range. Preserve natural phrase groups during word-by-word appearance; avoid making every character look like an independent button.
Correct words in their original positions
View HTML source
<!doctype html><html lang="zh-CN"><meta charset="utf-8"><meta name="viewport" content="width=device-width,initial-scale=1"><title>ElevenLabs|Scribe v2 · sequence</title><style>html,body{margin:0;width:100%;height:100%;overflow:hidden;background:#f7f6f0}canvas{display:block;width:100%;height:100%;object-fit:contain}</style><body data-mode="clip" data-start="23" data-end="34"><canvas width="1000" height="600"></canvas><script src="data.js"></script><script src="common.js"></script><script src="scene.js"></script><script src="bridge.js"></script></body></html>Show a confusable word first, then replace it in place with the correct proper term. Put prompts or vocabulary lists beside it and use a short connector to show influence; do not hide specific changes through wholesale paragraph reflow.
Match entity categories to words one by one
Mark names, organizations, or sensitive information with stable colors, highlighting the same item in the category table and body together. When masking information, affect only that field and preserve sentence context.
For another industry, prepare new vocabulary, speech content, and entity categories, retaining precise correspondence among word positions, time, and categories.
Sections and rhythm
Click a section to seek to it in the choreography example above.






Visual references · Discrete still frames
Creating other themes
- Adapt to structured-understanding products, preserving original input→correction→classification and location
- Actual capability evaluation requires independent data; demonstration figures are not detection results
Production constraints
Circular version mark / comparison bars / timeline / speaker paragraph / corrected word / entity label / connecting line
- Introduce measurement dimensions first, then explain capabilities through specific text changes
- Correct the word in the same position, removing the old before replacing it; entity colors must correspond to categories on the right
- use timeline and transcript-block positions and connectors to establish audio timing relationships