Kimi3Manim: concept to rendered film
Drive the Kimi3Manim six-agent pipeline (Concept Scout, Mathematical
Enricher, Visual Designer, Narrative Composer, Manim Coder, vision Render
Critic under a deterministic supervisor) to turn any math or physics
concept into a rendered 1080p Manim film.
When to Use
- The user asks to animate, visualize, or make a video/film of a
mathematical or physical concept, at any level from middle-school
geometry to graduate physics.
- The user wants to resume or debug a failed animation run.
- The user wants an existing Manim scene re-rendered.
Quick Reference
| Task |
Command (run in the repo directory) |
| Full pipeline, concept to mp4 |
uv run python -m k3_agents.supervisor "<concept>" --quality qh |
| Resume failed run at Stage 5 |
uv run python resume_pipeline.py [run_dir] |
| Re-render a scene, no model calls |
uv run manim -qh <file.py> <SceneClass> |
| One-time subscription login |
uv run kimi login |
| Install deps |
uv sync |
Quality flags: ql 480p (smoke test), qm 720p, qh 1080p, qk 4K.
--max-repairs N bounds coder/critic repair rounds (default 3).
Procedure
- Locate the repo (config
kimi3manim.repo, default ~/KimiK3Manim).
Missing? git clone https://github.com/HarleyCoops/KimiK3Manim.git,
then uv sync inside it. System deps: ffmpeg, texlive,
texlive-latex-extra, dvisvgm.
- Verify auth: either the user ran
uv run kimi login once (Kimi Code
OAuth subscription - the default), or MOONSHOT_API_KEY is set in
.env (also enables the vision critic stage).
- Warn the user that a full run takes minutes and spends Kimi quota
(six kimi-k3 calls at max reasoning plus rendering); never launch
runs in parallel.
- Phrase the concept the way the intended audience would ask it - the
Scout calibrates prerequisite depth from the phrasing. "why the
Pythagorean theorem is true" is as valid as "the double cover of
SO(3)".
- Run the supervisor command. The built-in preflight aborts with exact
instructions if the environment is broken - relay its message
verbatim and fix the environment; do not retry the pipeline.
- Outputs land in
output/k3_runs/<slug>/: numbered JSON artifacts
(01 knowledge graph ... 06 critique), generated scenes in scenes/,
per-attempt render_attemptN.log, and the mp4 under media/.
Deliver the mp4. [[as_document]]
Pitfalls
- Environment failures (missing manim/latex/pango) are NOT code bugs:
the supervisor aborts on them by design. Fix the environment; never
feed them to the coder repair loop.
- Re-running the same concept reuses its run directory - move it aside
first for a clean slate.
- Subscription-only auth cannot run the Stage 6 vision critic; the run
stops after rendering with the mp4 already on disk. Ship it and note
the critic was skipped, or set MOONSHOT_API_KEY to enable it.
- On WSL, keep the clone in the Linux filesystem (not /mnt/c) or uv and
renders run slowly;
export UV_LINK_MODE=copy silences the hardlink
warning.
Verification
- The run directory contains an mp4 under
media/ and artifacts 01-05.
- On failure, read the newest
render_attemptN.log in the run dir
BEFORE re-running; it contains the real traceback.
- Sample a few frames (
ffmpeg -ss <t> -i <mp4> -frames:v 1 f.png) and
confirm equations are legible and scenes match the visual spec before
presenting the film as final.
1---2name: kimi3manim3description: Turn math and physics concepts into rendered Manim films4license: MIT5---67# Kimi3Manim: concept to rendered film89Drive the Kimi3Manim six-agent pipeline (Concept Scout, Mathematical10Enricher, Visual Designer, Narrative Composer, Manim Coder, vision Render11Critic under a deterministic supervisor) to turn any math or physics12concept into a rendered 1080p Manim film.1314## When to Use1516- The user asks to animate, visualize, or make a video/film of a17 mathematical or physical concept, at any level from middle-school18 geometry to graduate physics.19- The user wants to resume or debug a failed animation run.20- The user wants an existing Manim scene re-rendered.2122## Quick Reference2324| Task | Command (run in the repo directory) |25|---|---|26| Full pipeline, concept to mp4 | `uv run python -m k3_agents.supervisor "<concept>" --quality qh` |27| Resume failed run at Stage 5 | `uv run python resume_pipeline.py [run_dir]` |28| Re-render a scene, no model calls | `uv run manim -qh <file.py> <SceneClass>` |29| One-time subscription login | `uv run kimi login` |30| Install deps | `uv sync` |3132Quality flags: ql 480p (smoke test), qm 720p, qh 1080p, qk 4K.33`--max-repairs N` bounds coder/critic repair rounds (default 3).3435## Procedure36371. Locate the repo (config `kimi3manim.repo`, default `~/KimiK3Manim`).38 Missing? `git clone https://github.com/HarleyCoops/KimiK3Manim.git`,39 then `uv sync` inside it. System deps: ffmpeg, texlive,40 texlive-latex-extra, dvisvgm.412. Verify auth: either the user ran `uv run kimi login` once (Kimi Code42 OAuth subscription - the default), or `MOONSHOT_API_KEY` is set in43 `.env` (also enables the vision critic stage).443. Warn the user that a full run takes minutes and spends Kimi quota45 (six kimi-k3 calls at max reasoning plus rendering); never launch46 runs in parallel.474. Phrase the concept the way the intended audience would ask it - the48 Scout calibrates prerequisite depth from the phrasing. "why the49 Pythagorean theorem is true" is as valid as "the double cover of50 SO(3)".515. Run the supervisor command. The built-in preflight aborts with exact52 instructions if the environment is broken - relay its message53 verbatim and fix the environment; do not retry the pipeline.546. Outputs land in `output/k3_runs/<slug>/`: numbered JSON artifacts55 (01 knowledge graph ... 06 critique), generated scenes in `scenes/`,56 per-attempt `render_attemptN.log`, and the mp4 under `media/`.57 Deliver the mp4. `[[as_document]]`5859## Pitfalls6061- Environment failures (missing manim/latex/pango) are NOT code bugs:62 the supervisor aborts on them by design. Fix the environment; never63 feed them to the coder repair loop.64- Re-running the same concept reuses its run directory - move it aside65 first for a clean slate.66- Subscription-only auth cannot run the Stage 6 vision critic; the run67 stops after rendering with the mp4 already on disk. Ship it and note68 the critic was skipped, or set MOONSHOT_API_KEY to enable it.69- On WSL, keep the clone in the Linux filesystem (not /mnt/c) or uv and70 renders run slowly; `export UV_LINK_MODE=copy` silences the hardlink71 warning.7273## Verification7475- The run directory contains an mp4 under `media/` and artifacts 01-05.76- On failure, read the newest `render_attemptN.log` in the run dir77 BEFORE re-running; it contains the real traceback.78- Sample a few frames (`ffmpeg -ss <t> -i <mp4> -frames:v 1 f.png`) and79 confirm equations are legible and scenes match the visual spec before80 presenting the film as final.
harleycoops