Skip to content

v0.1.18-alpha — first binary alpha

Pre-release
Pre-release

Choose a tag to compare

@cyscott cyscott released this 24 Jul 15:04
· 1 commit to main since this release

This is the first public playable binary alpha of Diddy Kong Racing Recompiled.

The download does not contain a ROM or extracted game assets. On first launch,
select your own legally obtained, unmodified Diddy Kong Racing USA Rev 1 ROM
(SHA-1 6d96743d46f8c0cd0edb0ec5600b003c89b93755).

Downloads

  • macOS Apple silicon: unzip the macOS archive and open
    Diddy Kong Racing Recompiled.app. The app is ad-hoc signed but not notarized,
    so macOS may require right-clicking the app and choosing Open the first time.
  • Steam Deck / Linux x86_64: extract the .tar.gz, then run the enclosed
    AppImage. The tar archive preserves its executable permission. It can also be
    added to Steam as a non-Steam game.
  • Verify either download with SHA256SUMS.

Tested

  • ROM selection and validation
  • Title and attract sequences, including the water scene
  • Xbox-compatible and Steam Deck controller input
  • Character selection
  • Normal races and close-up racer models, including Banjo and Krunch
  • Race completion and post-race screens
  • Adventure startup, overworld, save, and reload
  • Widescreen gameplay and 4:3 menu presentation

Alpha limitations

  • Full Adventure progression, multiplayer, bosses, and every course/vehicle
    combination have not been exhaustively validated.
  • Windows has not been validated and is not included in this release.
  • Expand widescreen can show a localized seam over one open-ocean area of the
    Adventure overworld. 4:3 presentation is unaffected.
  • Controller Pak emulation is not implemented; the game reports no Pak.

See Known Issues
and the Testing Guide
before filing a report.

Credits and development disclosure

This project builds on the work of the Diddy Kong Racing decompilation,
N64: Recompiled, N64 Modern Runtime, RT64, and their contributors. Development
was substantially assisted by OpenAI Codex, with direction and extensive
gameplay testing by the project maintainer. The repository includes a detailed
AI assistance disclosure.