Generate "Black Myth: Wukong " 3D assets in one sentence, Hu Yuanming's entrepreneurial project Meshy is new, free trial

The geometry is cleaner and more detailed, the workflow is more reasonable, Meshy's 3D generation capabilities have evolved again, and everyone can try it for free.

Man of destiny, where are you now?

There is no doubt that whether it is in the circle of friends or on various online platforms, the most popular game in the past two days is "Black Myth: Wukong".

Even if you don’t play games at all, you may have paid attention to this super masterpiece that is known as “creating the history of Chinese games” because of the record-breaking sales or the “tap water” that swept the entire network.

However, "Black Myth: Wukong" made its debut for the first time because of a live demonstration video. This trailer, which has nearly 60 million views, announces that in the game industry dominated by Western culture, Chinese players will soon have the heroic protagonist they dream of rooted in Chinese culture.

The eye-catching visual effects in the trailer are undoubtedly the direct reason for the popularity of "Black Myth: Wukong". Whether it is the careful design of each monster image or the exquisite sculptures and architecture in the temple, game science has achieved the ultimate in carving out visual effects.

If I want to transform Monkey Brother from such an original painting into a 3D object that can perform "72 transformations" in the game, we have to go through tedious processes such as modeling, rendering, and optimization.

                                                                              ^{From original painting to modeling}

Compared with characters, scene modeling is more complicated. Yang Qi, the art director of "Black Myth: Wukong", once said in an interview: Some scene architecture The prototype requires on-site inspection and research to form its own digital assets, and then it takes several months of processing and beautification before it can finally present such a beautiful picture. When scanning the first version of the Dazu Rock Carvings in Chongqing, the modeling once reached an exaggerated "1.2 billion faces", which shows the huge amount of work behind it.

Although the manual accuracy of this kind of workload is difficult to match, the modeling effect of AI tools is already impressive enough. It only takes 1 minute to get a modeling effect like the one below. The patterns on the armor are clear and distinct, and there are no signs of blurring or flickering on the edges.

You only need to select the sculpture style, then enter "Black Myth: Wukong" and the corresponding character. You do not need a high-configuration computer to directly get a 3D object close to the game style:

This 3D The AIGC tool comes from a startup called Meshy. The company's co-founder and CEO Hu Yuanming is a well-known scholar in computer graphics. He graduated from the Yao Class of Tsinghua University and has a PhD from MIT. He is also the author of the "TaiChi" programming language.

As of now, Meshy has been updated to the 4th generation. Since its launch in November 2023, Meshy has been at the forefront of the industry with its speed, fidelity, and ease of operation. Meshy 3 takes a step forward in improving the realism of Vincent 3D and the naturalness and delicacy of Tusheng 3D models. What new developments does Meshy 4 bring this time? Clean, detailed geometric mesh

In Meshy-4, Hu Yuanming’s team has broken through some technical difficulties in 3D AI generation. Whether you use text to 3D or image to 3D, you can now experience it The geometric quality of the generated models is greatly improved.

With an improved generation algorithm, Meshy-4 generates 3D models with very clean hard surfaces, without the bumps and dents that plagued the generation results in previous versions.

In addition, the geometric details generated by Meshy-4 have been greatly enhanced, allowing users to create highly complex models that capture nuances that previous AI-generated models could not capture.

The team also updated the “Discovery” page to match Meshy-4’s model quality enhancements. After the update, when you hover over a Meshy-4 generated model, you can instantly display its untextured version. This way, you can closely examine the underlying geometry of your model and see the clean hard surfaces and intricate details provided by the improved Meshy-4 algorithm.

New Text to 3D Workflow

The biggest update in the Meshy-4 user interface is the new Text to 3D workflow. In previous versions, the text to 3D process was divided into Coarse stage and Refine stage. While this approach was designed to incrementally improve model quality, it often left users feeling uncertain about the final results, and the Meshy team received a lot of user feedback about refined models deviating from crude results.

So, in Meshy-4, they split "text to 3D" into two steps: modeling and texturing in a more unique and focused way.

• Modeling phase: In this phase, Meshy-4 generates 4 textureless models based on text prompts. The mesh generated at this stage is clean and highly detailed, providing a solid foundation for the final asset to be generated.

• Texture Phase: After the modeling phase is complete, you can select a generated mesh result to continue generating textures. The texture was generated directly during the refinement stage of the previous version. The end result is visually very coherent.

Meshy team said that this new text to 3D workflow also paves the way for subsequent new features focused on modeling or texturing. The separation of the two allows users to more easily control the generation results, such as editing the mesh before generating textures or customizing multiple color variants for the same model.

If it doesn’t work once, try again

Some people say that the current AI generation is like a card drawing game, and it is difficult to know which time you can draw a good result. Therefore, the Meshy team added a "retry" function in the new version.

Even if your texture has been generated, "Retry" can still be used. However, it should be noted that after clicking "Retry", the previous results will be discarded, so it is best to be cautious.

Since retry involves more computing resources, this feature is set to be available only to subscribed users. For each model:

• Pro users have 4 retry opportunities.

• Max users have 8 retry opportunities.

• Max Unlimited users have unlimited retry opportunities.

Multiple models to choose from

In the previously released Meshy-3 Turbo, the team introduced a model selector for text-to-3D generation, which allows users to choose between Meshy-3 and Meshy- Switch between 3 Turbo algorithms. Meshy-4 extends this functionality further and you can select generative models in text-to-3D and image-to-3D tasks in the options below.

• Meshy-4: Meshy’s latest and most advanced model, capable of generating flat geometry, sharp angles and complex details.

• Meshy-3 Turbo: Meshy’s fastest model with less detail. Suitable for organic modeling or highly artistic styles.

• Meshy-3: Traditional model of Meshy, only suitable for text to 3D.

In addition, in the new version, the "Mode" option in image to 3D has been replaced. This is because it corresponds exactly to the underlying model: Organic Mode is powered by Meshy-3 Turbo, while Hard Surface Mode is powered by Meshy-4.

Hu Yuanming and the Meshy team

Meshy is a start-up company founded by Hu Yuanming, which mainly studies how to use AI to generate better 3D assets.

胡遠明はコンピューターグラフィックスの著名な学者であり、清華大学八尾クラスを卒業しました。

2019 年、MIT で博士課程の勉強中に、「Taichi」と呼ばれるオープンソースのコンピューター グラフィックス ライブラリの開発を主導し、コンピューター グラフィックスの分野で広く注目を集めました。 2021年3月にMITで電気工学とコンピュータサイエンスの博士論文の弁論に合格し、その後起業するために中国に戻り、「Tai Chi Graphics」という会社を設立した。同社は設立から1年も経たないうちに、5,000万ドルのシリーズAを含む2回の資金調達ラウンドを完了した。

2023年11月、胡源明氏の新しい起業家プロジェクトMesyが暴露されました。新会社が開発したオンライン生成AIツール（Meshedの初版）は、Alを使った3Dコンテンツ（モデル）をわずか1分で生成できる。その後数か月間、このツールは急速に改良され、現在 4 番目のメジャー バージョンが導入されました。

Meshy チームのメンバーは世界中におり、MIT、ハーバード大学、NVIDIA、Google、Meta などの有名な機関の出身であると報告されています。彼らは、直感的で簡単なコンテンツ作成方法を提供することで、現在の 3D コンテンツ制作エコシステムに革命を起こしたいと考えています。

しかし、ローマは一日にして成らずということも彼らは知っています。そのため、彼らは AI がアーティストやデザイナーの仕事をきっぱりと完全に置き換える未来を想像したことはありませんでした。彼らはまた、自分たちが構築するツールが創造性を高め、想像力の限界を押し広げるツールになることを望んでいます。

^{参考リンク：https://www.meshy.ai/zh/blog/meshy-4-break-grounds}

The above is the detailed content of Generate "Black Myth: Wukong " 3D assets in one sentence, Hu Yuanming's entrepreneurial project Meshy is new, free trial. For more information, please follow other related articles on the PHP Chinese website!