Are the commit hashes real?

No. They are 40-character hexadecimal strings produced by a seeded random generator. They look like real SHA-1 hashes but are not derived from any actual commit content.

What commit message style is used?

Conventional Commits, like feat(api): add pagination or fix: handle empty input. Types include feat, fix, chore, docs, refactor, test, perf, and style, with optional scopes.

How are the dates assigned?

Commits walk backwards from the current time, with each commit stepping back a random 1 hour to 3 days. This produces a believable descending chronology, just like a real git log.

What is this useful for?

Generating fixtures and demo data for changelog generators, git-history visualizers, release-note tools, and screenshots where you need realistic but non-sensitive commit data.

Fake Git Commit History Generator

Realistic git history without a repo

When you are building or demoing a changelog generator, a commit-graph visualizer, or release tooling, you need realistic git log output, but you do not want to expose a real repository. This tool synthesizes a believable commit history with proper hashes, authors, dates, and Conventional Commits messages, all generated locally.

How it works

A seed drives a mulberry32 pseudo-random generator so the same inputs always produce the same log.

Hashes are 40 hex characters, matching SHA-1 length. The oneline format shows the first 7, as git does by default.
Messages follow Conventional Commits: a type, optional (scope), then a colon and subject.
Dates start at “now” and step backwards a random interval per commit, so the history reads newest-first like real git log.

Full format looks like:

commit 3f9a2c...e1
Author: Alex Rivera <[email protected]>
Date:   Fri, 06 Jun 2026 10:12:00 GMT

    feat(api): add pagination to list endpoint

Tips and notes

Use the oneline format when you need a compact list for a changelog screenshot, and the full format when your tool parses author and date fields. Increase the commit count to stress-test rendering. Because everything is seeded, you can reproduce the exact same dataset in a test by reusing the seed.