News for hackers. Stuff that matters.

Show HN: FlashRecord – 2MB Python-native CLI screen recorder

https://github.com/Flamehaven/FlashRecord

Hi HN — I built FlashRecord, a tiny (≈2MB) Python-native CLI tool for screenshots and GIF recordings aimed at developers who want automation-friendly, scriptable screen capture without a GUI.

### What it is

- CLI-first and importable (import flashrecord) so you can plug it into scripts, tests, CI pipelines, or docs generation. - Outputs GIFs (and screenshots) with a pure-Pillow/NumPy implementation of a CWAM-inspired compression pipeline (multi-scale saliency, temporal subsampling, adaptive scaling). - Cross-platform (Windows/macOS/Linux), zero-config defaults, and production-ready with tests/docs.

---

### Why it might be interesting

- Tiny install and no heavyweight GUI/tooling to manage. - Designed for automation: generate evidence GIFs in CI, attach demo GIFs to PRs, or create tutorial assets from scripts. - Compression focuses on preserving visually important regions while reducing file size dramatically in typical UI demos.

---

Repo & license: https://github.com/Flamehaven/FlashRecord — MIT licensed.

---

I’m happy to answer technical questions, performance numbers, cross-platform quirks, or walk through the compression pipeline. Feedback, issues, and PRs welcome. What it is

CLI-first and importable (import flashrecord) so you can plug it into scripts, tests, CI pipelines, or docs generation.

Outputs GIFs (and screenshots) with a pure-Pillow/NumPy implementation of a CWAM-inspired compression pipeline (multi-scale saliency, temporal subsampling, adaptive scaling).

Cross-platform (Windows/macOS/Linux), zero-config defaults, and production-ready with tests/docs.

Why it might be interesting

Tiny install and no heavyweight GUI/tooling to manage.

Designed for automation: generate evidence GIFs in CI, attach demo GIFs to PRs, or create tutorial assets from scripts.

Compression focuses on preserving visually important regions while reducing file size dramatically in typical UI demos.

Quick try (from source)

git clone https://github.com/Flamehaven/FlashRecord cd FlashRecord pip install -e . flashrecord @sc # instant screenshot flashrecord @sv 5 10 # 5s GIF at 10 FPS (interactive by default)

Repo & license: https://github.com/Flamehaven/FlashRecord — MIT licensed.

I’m happy to answer technical questions, performance numbers, cross-platform quirks, or walk through the compression pipeline. Feedback, issues, and PRs welcome.

( 7 comments | 18 points | on HN )
Re:Show HN: FlashRecord – 2MB Python-native CLI screen recorder by butteredpecan on Sunday October 26, 2025 @02:34PM (#45714151)
A [vibecoded] warning might be appropriate.

What's the purpose of the AIPromptManager[0] class in a CLI screen recording project?

[0] https://github.com/Flamehaven/FlashRecord/blob/084e3308d7984...

Re:Show HN: FlashRecord – 2MB Python-native CLI screen recorder by whinvik on Sunday October 26, 2025 @04:03PM (#45714782)
What kind of usecases does a screenshot CLI have?
Re:Show HN: FlashRecord – 2MB Python-native CLI screen recorder by hnlmorg on Sunday October 26, 2025 @03:09PM (#45714399)
I’ve been using vhs recently. Same kind of solution: CLI, easy to install, bug focus on automation, supports gif, png, webp, etc. I’ve been really impressed with the output of vhs

https://github.com/charmbracelet/vhs

Re:Show HN: FlashRecord – 2MB Python-native CLI screen recorder by gus_massa on Sunday October 26, 2025 @03:04PM (#45714379)
Where is the license?
Re:Show HN: FlashRecord – 2MB Python-native CLI screen recorder by tecleandor on Sunday October 26, 2025 @03:08PM (#45714395)
Readme says "MIT License - See LICENSE file" bit there's no LICENSE file.

VISION.md file feels very weird to me...

Re:Show HN: FlashRecord – 2MB Python-native CLI screen recorder by bakugo on Sunday October 26, 2025 @03:12PM (#45714416)
The AI that generated the entire project forgot to add it.
Re:Show HN: FlashRecord – 2MB Python-native CLI screen recorder by hnlmorg on Sunday October 26, 2025 @03:18PM (#45714470)
There’s lots of other documents referenced in (for example) CONTRIBUTING that doesn’t exist in the repository.

I’ve got zero issue with using AI for development, but the author really should have read their documentation afterwards to sanity check it.

Though it’s also possible they’re not English literate. I also found some markdown documents written in Korean.