Spike
Nicolette-generated explainer videos for the AI-in-practice feed
Working target: 60–120 second videos that use original voiceover, diagrams, generated visuals and subtitles. No borrowed influencer footage, no unsupported claims, and only licensed or self-made media.
Recommended format
Repeatable short-form explainer
- Hook + problem framing in the first 10 seconds
- Voiceover from TTS with a neutral, practical tone
- Simple diagrams, charts, icons, and generated abstract visuals
- Burned-in subtitles plus a transcript for accessibility and reuse
- Source list in the post body or accompanying notes
Rights and brand safety
Commentary only
- Use only original content or licensed assets
- Do not reuse external footage unless explicitly licensed and documented
- Avoid claims that cannot be supported by a source
- Keep the voice and visuals clearly technical, not fake-human influencer style
Practical pipeline
Minimal toolchain
- Write a 120–180 word script
- Generate TTS audio
- Create still diagrams or a simple motion graphic strip
- Assemble with ffmpeg
- Add subtitles from the script or timing file
Publishing
Preview first
Host locally on the preview node for review, then embed or republish later only after the source list, licensing notes and claims have been checked.
Proof-of-concept command sketch
One-video prototype
tts → audio.wav
script + diagrams → frames/
ffmpeg -framerate 30 -i frames/%04d.png -i audio.wav \
-vf subtitles=captions.srt -c:v libx264 -pix_fmt yuv420p out.mp4
Acceptance criteria: the output is original, readable on mobile, includes captions, cites sources in the accompanying page, and can be regenerated from the same script + assets.
Operational guardrails
What this spike concludes
- Feasible without public-facing filming or stock-footage dependency
- Best built as a scripted assembly pipeline first
- Generated visuals are optional, not required, for a strong first prototype
- Safe publishing flow is preview → review → later platform upload
