A narration script can crowd a video before a voice is ever recorded. This exercise edits three original lines against the existing DreamForge coffee storyboard, calculates their word budgets and saves the decisions in an editable worksheet.
The source storyboard and exports are real text-planning outputs. The narration below is original written copy. We did not generate, record or listen to audio, so the timing calculations are planning budgets rather than measured speech durations or a voice-quality test.
Keep the shot and narration records connected
The worked storyboard has three stable shot IDs and duration suggestions of four, five and four seconds. Use those IDs in the script worksheet. Their combined 13 seconds is a draft edit target; it is not a rendering parameter that DreamForge has configured in a video provider.
For this worksheet, reserve half a second of each planned shot for silence. That leaves 3.5, 4.5 and 3.5 seconds as speech budgets. This is our editorial assumption for this example, not a recommended universal speaking rate. Real delivery may need more time, a different pause or a revised edit.
The intended narration adds a small sense of rhythm. It does not read out shot numbers, camera terms or everything already visible in the scene.
Inspect the before-and-after script
coffee-01: prepare
First draft: “The cup is on the table and the server is moving into position before coffee is poured.”
Edited script: A moment to get ready.
Decision: Remove a literal description of the arrangement; keep a short opening thought.
coffee-02: pour
First draft: “Now the coffee pours from the glass server into the cup while we hold a steady close-up view.”
Edited script: One slow pour, with room to pause.
Decision: Remove camera terminology and redundant object names from the spoken copy.
coffee-03: finish
First draft: “We finish by looking at the filled cup as a gentle trace of steam rises above the surface.”
Edited script: Then a quiet finish.
Decision: Give the ending less spoken information so the planned closing view has space.
The edits remove detail that the visual brief already carries and leave fewer spoken decisions to fit into each segment. That is the writing intent; only a real recording and viewing session could tell us whether the delivery and picture work together.
Calculate a budget, then test it with real speech
We counted English letter/number words, retaining internal apostrophes and hyphens as part of a word. Punctuation does not add a word. For each line:
budget words per minute = word count / (planned shot seconds - reserved pause seconds) × 60| Shot | Draft seconds / pause reserved | First draft words / budget wpm | Edited words / budget wpm |
|---|---|---|---|
| coffee-01 | 4 / 0.5 | 17 / 291.43 | 5 / 85.71 |
| coffee-02 | 5 / 0.5 | 18 / 240.00 | 7 / 93.33 |
| coffee-03 | 4 / 0.5 | 18 / 308.57 | 4 / 68.57 |
For the second line, seven edited words over 4.5 seconds gives 93.33 budget words per minute. The same slot had 18 words in the first draft, or 240 budget words per minute. This comparison identifies a crowded plan; it does not measure how a particular narrator or speech model performs.
The edited lines contain five, seven and four words. They have 16 words in total and 11.5 seconds of combined speech budget after the three illustrative pauses. There is no guarantee that a recording will last 11.5 seconds, because pronunciation, pauses and delivery are not determined by a word-count formula.
Keep pronunciation instructions out of the spoken copy
Put shot IDs such as coffee-02, aspect ratios such as 9:16, and timeline timecodes in the production sheet, not in the narration text. If an acronym or unusual name must be spoken, write down the intended reading for the authorized narrator and test that part first.
The worksheet is neither SSML nor an import into a voice service. ElevenLabs’ text-to-speech documentation describes model-specific prompting and controls; use the instructions for the exact service and model you select. Its current terms are a separate place to check permitted use. Do not assume a written pronunciation note configures a provider or grants a voice right.
Take the editable record into a recording session
- Narration CSV: first drafts, edited text, word counts, pause assumptions and calculated budgets.
- Readable Markdown worksheet: the same calculations and editorial decisions.
- Plain script TXT: copy the lines into your chosen writing or recording workflow.
- Complete planning handoff: the worksheets plus the native storyboard input and three export formats.
After recording authorized speech, measure each line including its actual pauses. If it overruns the planned slot, revise the text or shot timing before forcing the audio to fit. Listen for missing words, pronunciation, pace and abrupt changes between separately recorded sections. Then review narration, music, captions and the complete delivered file together.
The filled workflow checklist keeps those recording, listening and final-review stages marked Not run until they happen. A successfully written worksheet is useful evidence of planning; it is not evidence that a voice or final film has been approved.