
3Blue1Brown-style lesson on variational autoencoders
1.1K views · 4 likes · 8 saves · 0 reposts · Posted Sep 24, 2026 · 5:34 · Explainers
OUR NOTEA five-minute Manim explainer on VAEs with latent-space plots of real MNIST digits from trained models, cloned narration via Qwen3-TTS on MLX and original music, all from one short prompt.
Opus 5.5 can now make animated explainer videos end to end, in a style similar to 3Blue1Brown.
I gave it one short prompt.
A few minutes later, it produced a 5-minute video on VAEs with:
- Manim animations and math
- Models actually trained on MNIST, so every digit shown is real
- Cloned human narration using Qwen3-TTS on MLX
- Original background music
This changes how technical topics can be explained. A concept or research paper can go from a single prompt to a full video walkthrough.
Insane 🤯THE PROMPT
The creator didn't publish a prompt for this one. The post text is below.
MAKE ONE LIKE THIS
Three steps, using what the creators in this index used.- 1 · Copy the prompt
This creator didn't share a prompt. Start from our template instead.
PROMPT TEMPLATE → - 2 · Long piece? Give Claude a framework
- RavenMulti-agent research, coding and design orchestration★ 4.1k
- vox-directorPaper-collage style explainer and ad videos★ 2.1k
- video-talkcraftVoiceover-synced kinetic-text explainer videos★ 1.3k
- 3 · Render it to MP4
For a longer piece like this one, render with a framework so the timing is exact on every take.
CLAUDE CODE + HYPERFRAMES / REMOTION →
The creator didn't name their tools. Most videos like this are HTML, canvas or SVG animation written by Claude and screen-recorded or rendered to MP4 — how that works.
Indexed Sep 28, 2026, 8:28 AM · Views as of indexing. Video and post © their creator.