rendi-api/ffmpeg-cheatsheet

1. Repo type and boundary

- This is a documentation-first FFmpeg knowledge base, not an application or service.

- The architectural boundary is a curated set of Markdown assets that teach and standardize media-processing command patterns for humans and LLMs.

2. Top-level entry point

- README.md is the main landing page and acts like the product overview.

- It contains the full cheat sheet: concepts, categorized examples, caveats, and links to upstream FFmpeg references.

- For architecture purposes, treat this as the primary “index” into the knowledge system.

3. Skill module layout

- skills/ffmpeg/SKILL.md is a domain-specific skill package.

- The skills/ffmpeg folder indicates the repo is organized for agent consumption: a skill description plus supporting references.

- This suggests a clear separation between the broad cheat sheet and reusable skill-oriented guidance.

4. Reference layer

- skills/ffmpeg/references/command-patterns.md is a supporting reference artifact.

- Its role is likely to hold reusable command templates or canonical patterns, complementing the narrative examples in the README.

- This is the closest thing in the repo to a “shared library” of FFmpeg usage patterns.

5. Functional content domains

- The README is organized into major media-processing subsystems:

- simple editing and remuxing

- resizing, padding, and aspect-ratio management

- audio extraction, replacement, mixing, and fading

- advanced time-based editing and jump cuts

- overlays, subtitles, and compositing

- asset generation like thumbnails, GIFs, storyboards, and slideshows

- encoding settings and GPU acceleration

- These are not code modules, but they are the architectural domains of the knowledge base.

6. Core data flow model

- Most examples follow the same pipeline shape:

- one or more input media assets

- FFmpeg filters / filter graphs

- optional stream mapping

- encoding or stream copy

- a generated output file in a target container/codec

- Multi-input compositions use filter_complex and often split into parallel branches before recombining.

7. External dependencies and execution environment

- The repo assumes a local or scripted environment with the FFmpeg CLI and sometimes ffprobe.

- Sample commands reference remote assets on storage.rendi.dev, so the examples depend on network-accessible media.

- Some commands also depend on local assets like fonts, subtitle files, and images when using fontfile=, textfile=, or burned subtitles.

8. Key technical boundaries

- The README emphasizes the difference between stream copy/remuxing and re-encoding.

- It also distinguishes input seeking vs. output seeking, and when -c copy is safe or unsafe.

- These are important architectural boundaries because they determine whether a command is a low-cost container transform or a CPU-heavy transcoding pipeline.

9. Output and delivery concerns

- The examples cover production-oriented media outputs: MP4, MKV, MOV, WEBM, GIF, PNG/JPG thumbnails, and subtitle-bearing files.

- Encoding settings such as CRF, preset, yuv420p, faststart, and GPU encoders materially affect compatibility, startup behavior, and file size.

- In diagram form, these are the “delivery/packaging” concerns at the end of the media pipeline.

10. Audience and usage pattern

- The repo is designed for copy/paste execution and LLM-assisted exploration.

- Users are expected to browse the Markdown, search for a pattern, adapt a command, and run it locally.

- That makes the repo a knowledge repository with a strong “recipe” structure rather than a program with runtime control flow.

11. Operational style

- There is no evidence of build tooling, tests, or deployment infrastructure in the tree.

- The architecture is therefore intentionally lightweight: documentation, references, and executable command examples.

- The primary integration points are external FFmpeg docs, community references, and downloadable sample media, not internal services.

12. Best diagram framing

- Draw it as a documentation/skill system with:

- a root cheat sheet entry point

- a skill-specific FFmpeg module

- a small reference appendix

- external dependencies on FFmpeg, sample media hosts, and upstream documentation

- The main flow is “reference content -> adapted FFmpeg command -> local execution -> media output.”