Guides
Voice-over for courses and training that survives version 7
Courses are not narrated once. That is the whole argument.
Last checked 2026-08-14
In short
Narrate each slide as its own line, keep the voice and the pronunciation list fixed across the course, and store the script in a CSV beside the module. When a policy changes you re-render the three lines that changed, in the same voice, and the module is current again - which is the thing a human session cannot do a year later.
One line per slide
Keep the mapping between slide and audio file one-to-one and obvious - module3-slide14.mp3. It seems fussy until the compliance deadline moves and you need to change four slides in nine languages.
Consistency is the deliverable
- One voice per course, recorded in the CSV so a colleague can match it in a year.
- One pronunciation list for the organisation's product and policy names - it applies to every module on the workspace.
- One pace. Instructional narration wants deliberate; learners are also taking notes.
- One loudness target after export, so slide 14 does not shout.
Several languages, one master
Write the master in one language, translate into any of 35 and re-voice - including 23 Indian languages - keeping the slide mapping. Have a native speaker check each language once; the machine's mistakes are grammatical, not obvious.
Accessibility and procurement
Ship the SRT with the audio: captions are usually a requirement rather than a nicety, and they come out of the same render. For procurement, the answers to the questions you will be asked are written down - who processes the data, where it sits, what is stored and for how long - on the privacy, security and subprocessor pages, and the audio's synthetic origin is watermarked and logged.
Step by step
- Script per slide in a CSV Columns for slide id, text and voice. This file is the course's audio source of truth.
- Fix the vocabulary once Product names, policy names, acronyms into the pronunciation dictionary.
- Render the module as one job One file per line, named by slide.
- Translate if needed Same CSV, target language, same slide mapping.
- Re-render only what changes Next quarter's edit is three lines, not a studio day.
Questions
- Can I update one slide's narration later?
- That is the point of rendering per line: change the row, re-render that line, keep the voice identical.
- How many seats do I get?
- 3 on $15 a month, 10 on $39, with roles so a reviewer cannot spend the quota.
- Do you provide captions for the modules?
- Yes, an SRT timed from the render, in the same job.