Subtitling and Captioning

Core Philosophy

Subtitles must be read in real time while the viewer simultaneously watches the image and listens to audio. This triple constraint — reading speed, screen space, and synchronization — makes subtitling one of the most technically demanding forms of translation. Every subtitle must be timed precisely, short enough to read comfortably, and accurate enough to convey the speaker's meaning within severe character limits.

Key Techniques

• Timing and synchronization: Align subtitle appearance and disappearance with speech patterns.
• Condensation: Reduce spoken dialogue to essential meaning within character limits (typically 42 characters per line).
• Line breaking: Split subtitles at natural linguistic boundaries for comfortable reading.
• Reading speed calibration: Limit text to approximately 200 words per minute (or platform-specific guidelines).
• Speaker identification: Use positioning, colors, or labels to identify speakers in multi-speaker content.
• SDH conventions: Include sound descriptions and speaker IDs for deaf and hard-of-hearing viewers.

Best Practices

1. Two lines maximum per subtitle. Three lines obscure too much of the image.
2. Display each subtitle for at least 1 second and no more than 7 seconds.
3. Break lines at natural grammatical boundaries — between clauses, not mid-phrase.
4. Maintain subtitle timing with shot changes. A subtitle should not span a cut when avoidable.
5. Condense without changing meaning. Summarize what was said, do not add interpretation.
6. Use standard subtitle fonts (sans-serif, white with dark outline) for readability.
7. Leave minimum 2 frames gap between consecutive subtitles to signal text change.

Common Patterns

• Interlingual subtitling: Translation from source language to target language with condensation.
• Same-language captioning: Verbatim or near-verbatim transcription for accessibility.
• Forced narratives: Subtitles for on-screen text (signs, titles) in foreign-language content.
• Template-based workflow: Creating timed templates from source audio, then translating into them.

Anti-Patterns

• Subtitles that appear too briefly to read or linger too long after speech ends.
• Line breaks mid-word or mid-phrase that force the eye to scan unnaturally.
• Literal translation that produces subtitles too long to read at comfortable speed.
• Censoring or sanitizing dialogue beyond what the content creator intended.