Narrated Training Video Pipeline

Business impact

Made training content self-serve and repeatable: six narrated videos shipped in two days across three software rollouts, with no hand-recorded voiceovers and clickable links inside the video.

Problem

Slide-deck training wasn't self-serve, recording narration by hand didn't scale, and MP4 files can't carry clickable links to the tools being taught.

Solution

Engineered a scripted production pipeline: slides are rendered to images, narration is synthesized per slide with neural text-to-speech, segments are assembled with FFmpeg, and an HTML player overlays time-synced link hotspots on the video.

Results & highlights

• Produced and shipped six training videos in two days (2026-07-14 to 2026-07-15) across three software rollouts
• Placed hotspots automatically from slide-shape bounding boxes (python-pptx)
• Built in quality controls: pronunciation rules for acronyms and product names, and every segment validated with ffprobe before assembly

Competencies demonstrated

• Content automation: Scripted pipeline from slide deck and narration script to a finished, narrated MP4.
• AI text-to-speech: Neural text-to-speech per segment, with pronunciation rules for acronyms and product names.
• Media engineering: FFmpeg assembly, with every segment validated by ffprobe before it is used.
• Interactive learning: HTML player with time-synced, clickable hotspots placed automatically from slide-shape positions (python-pptx).
• Change management: Supported three software rollouts with ready-to-watch training.

Architecture

Notes

• Voice used: en-US-AndrewMultilingualNeural at -4% rate.