Decide what you need to build
AnyPS5 separates executable conversion from the libraries needed to run converted software. That distinction matters when deciding whether to try it: a successful relinker build would establish a conversion tool, not a working game runtime.
The current README describes a relinker that converts executables to native formats and replacement system PRX libraries for dynamic linking, without emulation or a separate runtime process. Those are the project's architectural claims. Our run did not verify converted-game execution.
What our recorded run establishes
Our run began on October 3, 2026, at 22:10:35.646 UTC, using commit ccd60ce0ca364ce26a0cccfb73223c9e2a1c4312. The recorded configuration was 3 CPUs, 8 GB of memory and image lab-cpp:1. The image label alone does not establish the operating-system or compiler version.
The harness labeled its first step install. It completed in 14 seconds, with download progress followed by CMake's configuration and generation messages. There is no installed-binary check in the supplied record. Calling this a successful CLI installation would overstate the evidence.
The build exited 1 after 44 seconds. At line 143 of the save-data implementation, the compiler rejected this expression:
std::chrono::clock_cast<std::chrono::system_clock>(written)
The diagnostic says that clock_cast is not a member of std::chrono. Ninja stopped after a subcommand failed. The supplied tail does not establish the compiler version or a verified remedy.
CTest then exited 8. The harness rounded its duration to 1 second; CTest's own tail reports 0.16 seconds of real test time. The recorded summary is 9 passed and 136 failed out of 145. However, the same tail marks many entries Not Run, including guest_addresses, guest_memory and strict_nid_filter. Other entries, such as elf_offsets and input_magic, are marked Failed.
These results describe an unsuccessful test command after an incomplete build. They do not establish 136 executed test failures or 136 independent product defects. The supplied counts have not been reconciled against a complete log to verify exhaustive unique coverage. No game conversion, game workload, frame-rate benchmark or current-version retest is recorded.
The repository scan recorded 10,465 files, approximately 2,300,908 source lines, 222 MB and 9 CI workflow files. It did not detect a Dockerfile or a top-level tests directory. That directory signal does not mean the project lacks tests: the recorded CTest run and build paths show tests elsewhere in the repository.
Current setup offers a smaller starting point
The build guide now documents a relinker-only build without submodules, SDL, Vulkan, FFmpeg, FreeType or the shader recompiler. It requires CMake, a C++20 compiler and a build tool. Python 3.9 or newer enables Python regression tests and must be discovered during configuration for those tests to be registered.
The guide says this mode builds on Linux, Windows and macOS, including Apple silicon. Its output is x86-64: Linux ELF by default, Windows PE with --windows, or Mach-O with --macos. The conversion host and execution host are separate considerations. Relinker-only mode does not build system libraries, and the guide says release libraries are supplied for Linux and Windows only. Converted games still need target-OS libraries and a compatible x86-64 host.
The full build requires x86-64, Git, CMake 3.22.1 or newer, Ninja, C++20 and recursive submodules. Linux uses GCC, G++ and binutils, with additional display and audio development dependencies described in the guide. FFmpeg binaries download during configuration unless FFMPEG_PREBUILT_DIR is supplied.
Windows full builds are no longer documented as limited to one package. The guide describes the required POSIX threads and exception support, naming WinLibs GCC 15.2.0 and MSYS2 UCRT64 GCC 16.2.0-6 as toolchain examples. It separately explains why Scoop's standalone GCC 15.2.0 cannot build the runtime. The technical-debt page still names the older specific MinGW-w64 GCC 15.2.0 requirement; follow the more detailed build guide when selecting a toolchain.
With the default BUILD_TESTING=OFF, the full build's default target does not produce the PRX libraries; the explicit libs target is required. On Windows, the guide also requires libgcc_s_seh-1.dll, libstdc++-6.dll and libwinpthread-1.dll beside the patched libraries. These instructions are source documentation, not steps we completed in a new run.
Compatibility needs a per-title answer
The README pinned to our measured commit reports Dreaming Sarah running at a stable 60 fps on a GTX 1050 Ti and i5-7500 3.4GHz. That is an upstream report, not our benchmark, and it does not establish compatibility for other games.
The current README directs readers to Game bring-up discussions for per-game status and test reports. The former compatibility-document URL was inaccessible in the supplied source collection. That does not prove its historical contents were false, but it prevents using that page to assert a current one-game total or current Windows and Linux results.
The README also defines its system-library progress percentage as the share of functions known to the project, not every PS5 system function. It is not a percentage of games that work.
The technical-debt document identifies concrete remaining uncertainty. Guest-module trampolines under --to-intel have synthetic relinker-test coverage but no verified game title on Linux or Windows. Some exports have unknown or assumed signatures, and several cancellation behaviors remain unverified. These are useful boundaries for research, not evidence that every title encounters the same failure.
How to choose your next step
For conversion research, start with the documented relinker-only mode, confirm which tests were registered and inspect the result on a compatible host. For runtime research, use a separate full-build directory, build the libraries and check a current report for the exact title you intend to investigate. We have not verified either path beyond the historical run described above.
For PS4 emulation, shadPS4 is a relevant alternative. Its README describes an early PS4 emulator and directs end users to QtLauncher. Kyty is relevant for studying an early emulator that reports simple PS4 games and PS5 homebrew. Neither description makes it a general replacement for AnyPS5's PS5 conversion work.
What the scores mean
These are editorial suitability ratings, not benchmark results. The scale is 1 for a substantial observed barrier, 2 for limited supporting evidence, 3 for a documented path with material verification gaps, 4 for detailed supporting documentation or validation within the assessed scope, and 5 for comprehensive evidence across that scope. Unknown adoption, compatibility and support quality earn no extra credit.
Setup receives 2/5 because our historical build stopped, while current documentation offers a smaller path we have not tested. Documentation receives 4/5 for its detailed separation of build modes, toolchains, library targets and troubleshooting, with a conflicting Windows requirement remaining in the debt page. Community receives 2/5 because the v0.1.1 release credits multiple contributors, but the supplied evidence does not establish support responsiveness or adoption. Maturity receives 2/5 because compatibility remains title-specific and the debt document records unverified runtime behavior. The v0.1.1 release is a historical source, not proof that it is the latest release.
The earlier AnyPS5 announcement covers the project's initial attention and compatibility report. This review answers the practical setup question: which component do you need, and what does the available evidence actually confirm?

