Establish the text source

Drama Skills uses the screenplay as the caption text source. Transcription can locate speech, but replay names, homophones and quiet lines. Do not silently treat recognition output as revised dialogue.

When audio says a wrong word or omits one, choose replacement media or an author-approved script revision with synchronization. Correct captions over wrong speech present two versions. A deliberate discrepancy needs an explicit creative design.

Generated text errors concern lettering inside generated images. This guide concerns captions added during editing.

Separate source, cut and sequence time

Locate speech in the source file, then subtract the cut’s in-point. Caption time is relative to this CUT. For the sequence, add preceding rendered durations.

Suppose the in-point is 3.00s, a complete line occupies 3.40–6.60s, and the out-point is 7.80s. This illustrates conversion, not an actual measurement report.

SpaceTimeMeaning
Source3.40–6.60sSpeech in the original file
Cut0.40–3.60sAfter subtracting the 3.00s in-point
Sequence2.40–5.60sIf preceding used material actually lasts 2.00s

Use final rendered segment durations for accumulation to avoid frame-rounding drift. A regenerated clip can change speech position even with unchanged words; replay affected lines.

Original example: leave the handoff visible

In the sample script, Alan places a blue key ring in a neighbor’s palm: “This opens the side door. Put it back in the blue box after closing.” The neighbor pockets it and nods. The handoff and box are established story information.

A single central caption covers both hand and box. Locate real speech and pauses, display readable successive captions, then check hands, box and interface space. Do not change “side door” to “front door” for layout or invent “tell nobody.”

If lower placement still masks the hand, compare moving text, adjusting framing or using another approved shot. This version’s built-in captions use a uniform style without per-CUT placement instructions. Use an external editor or subtitle layer for complex avoidance and inspect the output.

Review the finished picture and sound

Replay each line against text, inspect frames for size, order, split key words and occlusion, then watch at normal speed. Long-line splitting and caption styling are project/tool choices, not universal short-drama rules.

Treat errors by cause: check words against script and audio; remeasure timing and in-points; change layout or picture for occlusion; inspect a cut for a missing line. Do not carry old alignment findings onto new edits or regenerated footage.

Hard captions, separate subtitles or none can follow the work’s needs while honoring agreed delivery requirements. See the edit-list template.

FAQ

Can a correct transcript serve directly as captions?

Still verify script text, actual speech and edited timing. Accurate recognition does not establish readable timing or unobstructed placement.

Do it with the skill

Use $short-drama-edit with existing footage and screenplay. Preserve line text, locate speech and convert to CUT-relative times; realign after regeneration or changed in-points. Handle complex placement externally and review actual output audio and frames.

Read the method: Audio seams, captions and enacted mixing · Choose cuts by information and performance

About Drama SkillsThe short-drama-edit skill on GitHub

All guides