Skip to content

ForgePlay 2.0.0 (Build 6)

Latest

Choose a tag to compare

@Facta-Leopard Facta-Leopard released this 26 Sep 18:21

ForgePlay 2.0.0 (Build 6)

한국어 · English

한국어

ForgePlay 2.0.0은 macOS 27 Golden Gate를 기반으로 그래픽 기능과 게임 실행 플랫폼을 확장한 대규모 업데이트입니다.

HyPER-GAN 가중치를 활용한 DLSS5 Emulation(베타), 새로운 프레임 보간 방식, 독립적으로 사용할 수 있는 Frame Check를 제공합니다. Battle.net·STOVE·Epic Games 실행 환경을 추가하고, Rosetta 실행 호환성, 복구 스냅샷 관리와 AI 문제 진단도 개선했습니다.

업데이트 내용

macOS 27 Golden Gate 지원

  • ForgePlay 2.0.0부터 Apple Silicon Mac과 macOS 27 Golden Gate 이상이 필요합니다.
  • Apple Silicon 환경에 맞춰 앱의 실행 구성을 정비했습니다.
  • Intel Mac은 지원하지 않습니다. 내부 Windows 실행 엔진에 필요한 Rosetta 지원은 유지됩니다.

Rosetta 실행 환경 개선

  • Rosetta 환경에서 Windows 프로그램을 실행할 때의 호환성과 초기 환경 확인 절차를 보완했습니다.
  • Rosetta 설치 여부를 실제 Intel 코드 실행 가능 여부로 확인합니다.
  • Rosetta가 없으면 앱의 Rosetta 설치 버튼으로 Apple의 공식 설치 절차를 시작할 수 있습니다.
  • 약관 동의나 관리자 인증은 Apple 안내에 따라 직접 진행하며, 설치 후 ForgePlay에서 상태를 다시 확인할 수 있습니다.

DLSS5 Emulation(베타)

세계 최초*로, NVIDIA DLSS 코드·가중치에 의존하지 않는 HyPER-GAN 기반 실사화 후처리를 macOS에서 실행 중인 Windows 게임의 완료된 화면 프레임에 실시간으로 통합했습니다.

공개된 HyPER-GAN의 사전 학습 가중치를 활용해 DLSS 5와 유사한 실사화 효과를 근사하는 독자적인 파이프라인입니다. 새로운 모델을 발명한 것이 아니라, 기존 모델을 Mac의 Windows 게임 실행·표시 경로에 연결한 구현입니다.

ForgePlay는 2026년 9월 19일, 이미 MVP 테스트를 마친 상태에서 전후 비교 결과를 공개했습니다. 이번 버전은 그 MVP를 바탕으로 후속 개선과 제품 통합을 진행한 결과입니다. 당시 MVP 결과 공개 기록

  • Core AI와 Metal을 연동하며, AI 추론은 Neural Engine(NPU)을 우선 사용하도록 구성했습니다.
  • 게임 화면의 색상·밝기·대비를 보정하고, 원본 색조를 유지하면서 녹색 편향을 줄이는 처리를 적용했습니다.
  • 화면 스타일, 보정 강도와 보정 반복 횟수 1~5회를 선택할 수 있습니다.
  • 프레임 생성과 함께 사용할 수 있으며, 두 기능은 각각 켜고 끌 수 있습니다.
  • 새로운 설정에서는 DLSS5와 프레임 생성이 기본으로 꺼져 있습니다. 기존에 저장한 설정은 유지됩니다.

NVIDIA DLSS 5를 포팅한 기능이 아니며, NVIDIA의 공식 모델이나 가중치를 사용하지 않습니다.

활용한 HyPER-GAN은 MIT 라이선스에 따라 사용합니다. 출처와 라이선스 안내는 앱의 DLSS5 Emulation 옵션 옆 ⓘ 설명과 앱에 포함된 라이선스 문서에서 확인할 수 있습니다.

DLSS5 Emulation 권장 사양

AI 연산 성능과 메모리 대역폭을 고려해 다음 구성을 권장합니다.

  • M1~M4 세대: Max 칩 이상
  • M5 세대: Pro 풀칩 구성 이상

이는 ForgePlay의 설치 최소 사양이 아니라 DLSS5 Emulation 사용을 위한 개발자 권장 기준입니다. 실제 성능은 게임·해상도·설정에 따라 달라지며, 모든 게임에서 특정 FPS를 보장하지는 않습니다.

개발과 실기기 확인은 M4 Pro 환경을 중심으로 진행했습니다. M5 계열을 비롯한 다른 기기에서의 실행 결과도 보내주시면 최적화에 큰 도움이 됩니다.

속도가 느리다면

720p · 속도 우선 옵션을 사용해 보세요. 화면 비율을 유지하면서 내부 보정 해상도를 낮춘 뒤, 원래 표시 크기로 확대합니다. 새 설정에서는 이 옵션이 기본으로 선택됩니다.

보정 해상도와 반복 횟수가 높을수록 연산량과 메모리 부담이 커집니다. 프레임 저하나 입력 지연이 느껴진다면 반복 횟수를 낮추거나, 기능을 끈 상태와 비교해 주세요.

* ‘세계 최초’의 범위: NVIDIA DLSS 코드·가중치에 의존하지 않는 HyPER-GAN 기반 실사화 후처리를, macOS에서 실행 중인 Windows 게임의 완료 프레임에 실시간 통합했다는 범위에 대한 ForgePlay 개발자의 주장입니다. 2026년 9월 27일 기준 공개자료 조사에서는 같은 범위의 선행 적용 사례를 확인하지 못했습니다. 전 세계의 미공개·미색인 구현까지 인증한 결과는 아니며, HyPER-GAN 모델 자체의 최초 개발을 주장하는 것은 아닙니다.

프레임 생성 기능 확장 및 개선

기존의 기본 혼합 방식에 세 가지 움직임 기반 보간 방식을 추가했습니다. 총 네 가지 방식 중에서 선택할 수 있습니다.

  • Simple — 기본 혼합: 두 화면을 혼합하는 방식입니다.
  • Motion Lite — 움직임 속도 우선: 처리 부담을 줄이면서 움직임을 추정합니다.
  • Motion Quality — 정밀 추정: 움직임을 더 촘촘하게 추정합니다.
  • Motion Repair — 경계 복원: 정밀 추정 후 빈틈이나 어긋난 경계를 보완합니다.

새로운 보간 방식은 베타 기능입니다. 게임과 기기에 따라 효과나 처리 비용이 다르므로 비교해 선택해 주세요.

또한 이전 버전에서 프레임 생성 사용 시 sRGB 화면이 과도하게 밝아지거나 희게 뜨던 색상 처리 문제를 수정했습니다.

Frame Check 기능 강화

프레임 생성이나 DLSS5를 사용하지 않아도 Frame Check를 독립적으로 사용할 수 있습니다.

화면 우측 상단에서 다음 정보를 확인할 수 있습니다.

  • Original: 게임의 원본 프레임 제출 속도
  • DLSS5: AI 화면 보정 처리 속도
  • FG: 실제 표시된 생성 프레임 수
  • Displayed: 최종 화면 표시 FPS

각 수치는 서로 다른 처리 단계를 나타냅니다. Original과 DLSS5를 더한 값이 최종 FPS는 아닙니다. 실제 표시 시점을 확인할 수 없는 환경에서는 Displayed가 --로 표시될 수 있습니다.

WineD3D의 OpenGL 경로 정비

WineD3D의 렌더링 경로를 OpenGL 기반으로 정비했습니다.

L4D2(Left 4 Dead 2)에서 화면이 희게 뜨거나 일부 그래픽이 깨지던 문제에 대응하고, 관련 색상·화면 표시 처리를 개선했습니다. 특정 게임에만 적용되는 예외가 아니라 WineD3D의 공통 실행 경로에 반영했습니다.

이 경로에서는 Vulkan·MoltenVK 대신 OpenGL을 사용합니다. 다만 DXVK 등 다른 백엔드에서 사용하는 Vulkan·MoltenVK까지 제거한 것은 아닙니다.

쉽게 비유하면, Vulkan은 DirectX 12처럼 개발자가 낮은 수준의 처리를 직접 제어하는 방식에 가깝고, OpenGL은 DirectX 11처럼 드라이버가 상대적으로 많은 처리를 맡는 방식에 가깝습니다. 설계 방식을 이해하기 위한 비유이며, 기능이나 성능이 일대일로 대응한다는 뜻은 아닙니다.

프리픽스 복구 스냅샷 자동 정리

실행 환경을 갱신하거나 복구할 때 만들어지는 백업 사본과 임시 작업본의 정리를 개선했습니다.

  • 자동 작업이 성공적으로 끝나면 불필요해진 복구 사본을 정리합니다.
  • 실패·중단된 작업의 복구 자료와 사용자가 직접 만든 백업은 보존합니다.
  • 남아 있는 사본은 문제 진단 → 복구 스냅샷 관리에서 확인하고 정리할 수 있습니다.

‘프리픽스’는 Windows용 프로그램과 설정을 보관하는 실행 환경 폴더입니다. 백업을 삭제하면 해당 시점으로 되돌릴 수 없으므로, 내용을 확인한 뒤 필요한 항목만 정리해 주세요.

게임 플랫폼 확장(베타)

Steam 외에 다음 플랫폼의 설치·실행 화면을 추가했습니다.

플랫폼 현재 확인한 범위
Battle.net 런처 실행 및 브라우저 로그인
STOVE 런처 실행 및 브라우저 로그인
Epic Games 설치 및 런처 업데이트 진입

런처 실행이나 로그인 확인이 해당 플랫폼의 모든 게임 실행을 보장하는 것은 아닙니다.

Battle.net의 Safari 로그인 연결에 필요한 경우, 앱의 안내를 확인한 뒤 연결 준비를 진행해 주세요. 사용자 동의가 필요한 시스템 절차는 직접 확인할 수 있도록 구성했습니다.

관련 기능의 문제와 게임별 실행 결과를 알려주세요.

AI 문제 진단 개선

macOS 27 Golden Gate의 개선된 Apple Foundation Models 기반 로컬 AI를 활용하도록 문제 진단을 정비했습니다.

  • 실행 로그와 진단 정보를 바탕으로 가능한 원인과 확인할 해결 방법을 안내합니다.
  • 채팅으로 사용법과 실행 문제를 이어서 질문할 수 있습니다.
  • 지원 기기에서 확대된 모델의 문맥 한도를 활용하고, 실제 사용 가능한 범위에 맞춰 진단 정보를 구성합니다.
  • 진단 기록과 복구 스냅샷을 함께 관리할 수 있습니다.

AI 진단을 사용하려면 지원 기기에서 Apple Intelligence가 활성화되어 있고 모델 준비가 완료되어 있어야 합니다. 진단 결과는 참고 안내이며, 실제 원인이나 해결을 보장하지는 않습니다.

앞으로는 게임 호환성 제보와 확인된 해결 사례를 바탕으로 진단 자료를 보강하고, 검색 기반 정보 활용(RAG) 등을 통해 정확성을 높여갈 계획입니다.

ForgePlay 런처(베타)

Mac·VR·고전 게임 등으로 기능을 확장하기 위한 새로운 ForgePlay 런처를 추가했습니다.

현재는 Mac 실행 화면으로 연결되며, VR과 고전 게임 영역은 개발 중입니다. 앞으로 게임 실행 보조 기능과 별도 후처리 기능을 확장할 기반으로 활용할 예정입니다.

FPS 게임용 확대경과 같은 보조 기능도 검토하고 있으며, 이번 버전에 포함된 기능은 아닙니다.

스플래시 이미지 변경

사용자 제안을 반영해, 시작 화면을 기존 대장장이 토끼 이미지에서 ForgePlay 앱 아이콘 중심의 이미지로 변경했습니다.

설치 정보

  • 지원 환경: Apple Silicon Mac, macOS 27 Golden Gate 이상
  • 설치 파일: ForgePlay-2.0.0-6.dmg
  • 앱과 DMG는 Apple 공증을 완료했습니다.

실행 중인 ForgePlay와 Windows용 프로그램을 종료한 뒤, DMG 안의 ForgePlay.app 전체를 응용 프로그램 폴더로 복사해 주세요.

실행 결과와 피드백을 부탁드립니다

잘 되는 게임도, 안 되는 게임도 모두 도움이 됩니다.

GitHub Issues에 다음 정보를 함께 알려주세요.

  • ForgePlay 버전과 Mac 칩셋·메모리
  • 게임명과 사용한 플랫폼
  • 선택한 그래픽 백엔드
  • DLSS5 및 프레임 생성 사용 여부와 설정
  • 문제가 발생한 상황과 Frame Check 수치

여러분의 실제 사용 경험을 바탕으로 호환성과 성능을 계속 개선하겠습니다. ForgePlay를 응원해 주셔서 감사합니다.


English

ForgePlay 2.0.0 is a major update that expands graphics features and supported game platforms on macOS 27 Golden Gate.

It introduces DLSS5 Emulation (Beta) using HyPER-GAN weights, new frame-interpolation methods, and an independently usable Frame Check. It also adds launch environments for Battle.net, STOVE, and Epic Games, alongside improvements to Rosetta compatibility, recovery-snapshot management, and AI diagnostics.

What's new

macOS 27 Golden Gate support

  • Starting with ForgePlay 2.0.0, an Apple Silicon Mac running macOS 27 Golden Gate or later is required.
  • The app's launch configuration has been refined for Apple Silicon.
  • Intel Macs are not supported. Rosetta support remains in place for the internal Windows runtime.

Rosetta launch-environment improvements

  • Improved Windows-program compatibility under Rosetta and the initial environment checks.
  • Rosetta availability is checked by testing whether Intel code can actually run.
  • If Rosetta is missing, the app's Install Rosetta button opens Apple's official installation procedure.
  • You complete any license agreement or administrator authorization directly through Apple's prompts. You can then check the status again in ForgePlay.

DLSS5 Emulation (Beta)

In a world first, ForgePlay integrates real-time HyPER-GAN-based photorealistic post-processing into the completed frames of Windows games running on macOS, without relying on NVIDIA DLSS code or weights.*

This independently developed pipeline uses publicly available pretrained HyPER-GAN weights to approximate a DLSS 5-like photorealistic effect. The contribution is not the invention of a new model, but the integration of an existing model into the Windows-game execution and presentation path on Mac.

On September 19, 2026, ForgePlay published before-and-after results after completing MVP testing. This release builds on that MVP with further improvements and product integration. Original MVP-results publication record

  • Connects Core AI with Metal, with AI inference configured to prefer the Neural Engine (NPU).
  • Adjusts color, brightness, and contrast, with processing that preserves the original color tones and reduces green bias.
  • Offers a choice of visual style, adjustment strength, and 1–5 processing passes.
  • Can run alongside frame generation; each feature can be enabled or disabled independently.
  • DLSS5 and frame generation are off by default for new settings. Previously saved choices are preserved.

This is not a port of NVIDIA DLSS 5 and does not use NVIDIA's official models or weights.

HyPER-GAN is used under its MIT license. Attribution and licensing information are available in the ⓘ help beside DLSS5 Emulation and in the license documents included with the app.

Recommended hardware for DLSS5 Emulation

Considering AI compute performance and memory bandwidth, we recommend:

  • M1–M4 generations: a Max chip or higher
  • M5 generation: a full-configuration Pro chip or higher

These are the developer's recommendations for using DLSS5 Emulation, not ForgePlay's minimum installation requirements. Actual performance varies by game, resolution, and settings; a particular frame rate is not guaranteed across all games.

Development and on-device checks have primarily used an M4 Pro. Reports from other devices, including the M5 family, are valuable for further optimization.

If performance is slow

Try 720p · Speed Priority. It reduces the internal processing resolution while preserving the aspect ratio, then scales the result back to the original display size. This option is selected by default for new settings.

Higher processing resolutions and more passes increase compute and memory demands. If you notice lower frame rates or input latency, reduce the pass count or compare with the feature disabled.

* Scope of the “world first” claim: This is the ForgePlay developer's claim concerning the real-time integration of HyPER-GAN-based photorealistic post-processing into completed frames of Windows games running on macOS, without relying on NVIDIA DLSS code or weights. A review of publicly available material as of September 27, 2026 did not identify an earlier implementation within that same scope. It is not a certification covering all unpublished or unindexed implementations worldwide, nor a claim to have originally developed HyPER-GAN itself.

Expanded and improved frame generation

Added three motion-based interpolation methods alongside the existing simple-blending method. Four methods are now available:

  • Simple — basic blending: Blends two frames.
  • Motion Lite — speed-focused motion estimation: Estimates motion with lower processing overhead.
  • Motion Quality — precise motion estimation: Estimates motion in greater detail.
  • Motion Repair — boundary repair: Builds on precise estimation to repair gaps and misaligned boundaries.

The new interpolation methods are beta features. Their effectiveness and processing cost vary by game and device, so try them and choose the one that works best for your setup.

Also fixed color processing that could make sRGB images excessively bright or washed out when frame generation was enabled in the previous version.

Enhanced Frame Check

Frame Check can be used independently, even with frame generation and DLSS5 disabled.

The upper-right overlay shows:

  • Original: The game's original-frame submission rate
  • DLSS5: The AI image-processing rate
  • FG: Generated frames actually presented
  • Displayed: Final presentation FPS

These values describe different processing stages. Adding Original and DLSS5 does not give the final FPS. Displayed may show -- when actual presentation timing is unavailable.

Refined WineD3D OpenGL path

WineD3D's rendering path has been reworked around OpenGL.

This addresses washed-out images and graphical corruption encountered in L4D2 (Left 4 Dead 2), with improvements to color processing and presentation. The changes apply to WineD3D's common execution path, not a game-specific exception.

This path uses OpenGL instead of Vulkan/MoltenVK. Vulkan and MoltenVK have not been removed from other backends, such as DXVK.

As a simple analogy, Vulkan is closer to DirectX 12 in giving developers direct control over lower-level operations, while OpenGL is closer to DirectX 11 in leaving comparatively more work to the driver. This compares their general design approach, not a one-to-one equivalence in features or performance.

Automatic cleanup of prefix recovery snapshots

Improved cleanup of backup copies and temporary working copies created during environment updates or recovery.

  • Unneeded recovery copies are cleaned up after an automatic operation finishes successfully.
  • Recovery material from failed or interrupted operations, and backups created manually by the user, are preserved.
  • Remaining copies can be inspected and managed under Diagnostics → Recovery Snapshots.

A “prefix” is the environment folder that holds Windows programs and their settings. Deleting a backup removes the ability to return to that point, so check its contents before cleaning it up.

More game platforms (Beta)

Added installation and launch screens for the following platforms alongside Steam:

Platform What has been verified so far
Battle.net Launcher startup and browser login
STOVE Launcher startup and browser login
Epic Games Installation and entry into the launcher-update process

Successful launcher startup or login does not guarantee that every game on that platform will run.

If Battle.net requires preparation for Safari login, follow the app's instructions to set up the connection. System procedures requiring your consent are presented for you to review directly.

Please report any issues with these features and the results of individual games.

Improved AI diagnostics

Refined diagnostics to use the improved local AI powered by Apple Foundation Models on macOS 27 Golden Gate.

  • Uses execution logs and diagnostic information to suggest possible causes and troubleshooting steps.
  • Supports follow-up questions about usage and launch problems through chat.
  • Uses the expanded model context available on supported devices and prepares diagnostic information within the actual available limit.
  • Lets you manage diagnostic history and recovery snapshots together.

AI diagnostics require Apple Intelligence to be enabled and its model to be ready on a supported device. Diagnostic results are guidance, not a guarantee of the cause or a successful fix.

We plan to expand diagnostic material using game-compatibility reports and confirmed solutions, and improve accuracy through retrieval-augmented generation (RAG) and related approaches.

ForgePlay Launcher (Beta)

Added a new ForgePlay launcher as a foundation for expanding into Mac, VR, classic games, and more.

It currently opens the Mac launch interface. The VR and classic-game areas are in development. The launcher will provide a foundation for additional game-launch assistance and post-processing features.

We are also considering assistance features such as a magnifier for first-person shooters. These are not included in this release.

Updated splash image

Following user feedback, the startup screen now uses an image centered on the ForgePlay app icon instead of the previous blacksmith rabbit.

Installation

  • Requirements: Apple Silicon Mac, macOS 27 Golden Gate or later
  • Installer: ForgePlay-2.0.0-6.dmg
  • The app and DMG have completed Apple notarization.

Quit ForgePlay and any running Windows programs, then copy the entire ForgePlay.app from the DMG into Applications.

Share your results and feedback

Reports of both working and non-working games are helpful.

Please include the following when reporting through GitHub Issues:

  • ForgePlay version, Mac chip, and memory
  • Game title and platform
  • Selected graphics backend
  • Whether DLSS5 and frame generation were enabled, and their settings
  • What happened and the Frame Check readings

Your real-world experience helps us keep improving compatibility and performance. Thank you for supporting ForgePlay.