news

한어Русский языкEnglishFrançaisIndonesianSanskrit日本語DeutschPortuguêsΕλληνικάespañolItalianoSuomalainenLatina

Text: Zhou Ke, Sailing to the Sea

Although it has been around for nearly 10 years, Honor of Kings still earns over a billion yuan a month. Every game manufacturer hopes to create the next "king", but have you ever thought that the road to the next "king" may be in the hands of ordinary people?

Is it difficult to make games? There are barriers from R&D to promotion, but the integration of AI technology may open a door for ordinary people in this field. "AI+3D" points the way for the birth of large 3D models.

On July 16, Tencent's large-scale model application "Tencent Yuanbao" launched the "3D Character Dream Factory" gameplay. By opening the Tencent Yuanbao APP, you can upload a clear character portrait and then generate a 3D character.

3D Character DreamWorks provides multiple templates for users to choose from. After the generation, the 3D character can be viewed in 360 degrees, and a 3D printing link can be generated to actually print out the character.

Different from large models generated from text, pictures and videos, 3D large model products can truly realize the transition from virtual to reality.

However, the application of large 3D models goes far beyond this. Tripo, a product of VAST, a company specializing in large 3D model products, was launched last year. Its 3D model generation tool has accumulated hundreds of thousands of users and generated more than 4 million 3D models.

Compared with drawing, writing and shooting videos, the technical threshold of 3D generation is relatively high for most people. However, the birth of large 3D models has made it easy to create 3D models, and the market door has been opened to more ordinary users. Next, which industries will usher in disruptive changes?

01. With the support of AI, 3D large models enter the fast lane

When it comes to 3D, the first application scenario that people think of is games. To complete the design of a 3D game character, multiple processes are required, from original painting setting to model building to animation production. The procedure is relatively complicated and requires a high accumulation of 3D modeling technology.

However, after trying Tripo, we found that creating a 3D image is not difficult.

On the Tripo homepage, enter the prompt words and you can generate the 3D model you want. The process is as simple as generating pictures and videos.

We gave a prompt word and asked it to generate "a cute panda wearing a yellow hat and carrying a schoolbag walking", and Tripo generated images of 4 pandas for us.

Select one of them and optimize it.

Objectively speaking, when using Tripo, you can feel that its response speed is relatively fast. After receiving the prompt word command, the 3D image is generated in just a few tens of seconds.

Taking the word "panda" as an example, Tripo can accurately display the image of the panda, the yellow hat, the blue schoolbag, and the walking action, and it scores full marks in the display of the word elements.

However, when you enlarge the model, you can see that the image of the panda still has some flaws. For example, the lines of the panda's body are not smooth enough, the edges of the backpack are not finely processed enough, etc.

From the perspective of an ordinary user, without any accumulation of 3D modeling technology, using Tripo to generate a panda image of such quality is already a very good effect.

The company behind Tripo is VAST. Its CEO Song Yachen said in an interview: "The maturity of the 3D big model has surpassed the level of GPT3 or Midjourney V3. By September 2024, we believe it can reach the level of GPT3.5 or Midjoueney V4."

Song Yachen emphasized that "enriching 3D content modalities and lowering the threshold for 3D content creation" has been well reflected in Tripo's products. At present, Tripo's 3D generation effects still have a lot of room for upward expansion, which needs to be achieved through technological accumulation and iteration.

In overseas markets, the development of 3D modeling AI-assisted tools is also very rich. For example, another tool we have paid attention to, Masterpiece X, uses generative artificial intelligence technology to allow users to quickly create 3D models containing meshes, textures, and animations by entering simple text prompts.

Like Tripo, Masterpiece X is very friendly to users without modeling expertise. The threshold is very low and users only need to enter prompt words.

Masterpiece X supports the generation of various types of 3D models, including people, animals, and objects. The generated models can be downloaded and used unlimited times, and like Tripo, users can adjust the details of the generated 3D models.

However, in actual use, we found that the effects and generation speed of Masterpiece X were not outstanding.

We gave the same prompt "A lovely panda, wearing a yellow hat and carrying a blue schoolbag, is walking", and Masterpiece X took 6 minutes to generate the first 3D model.

The final generated 3D model is a texture effect, which is different from Tripo.

And the performance of elements such as panda, backpack, hat, and walking is not very good.

The tips we give are from the perspective of a user with no modeling experience at all, using Tripo and Masterpiece X to generate 3D models simultaneously.

Masterpiece X is not free. Each generation requires 50 points. However, users can earn points by participating in various activities in the community to continue using it.

In terms of effects, Masterpiece X also stated that it is not ready for high-fidelity effects and there is no way to precisely control every detail.

From a commercial perspective, it is obvious that Tripo's generation effect is better and the threshold for ordinary users is lower.

In addition to Masterpiece X and Tripo, there are some other 3D generation models that are attracting much attention on the market, such as Nextech3D.AI, Spline, 3DFY.ai, etc. These 3D model generation products have not been around for long, and the technology is not very sophisticated. Some applications cannot even achieve the presentation effects that can be used directly for commercial purposes.

On the one hand, this shows that the technical requirements for 3D model generation are relatively high. On the other hand, it also shows that there is still a lot of room for exploration in this market in the future, and its future commercial opportunities and scenarios are worthy of industry attention.

By paying attention to products such as Nextech3D.AI and Tripo, we found that in addition to the familiar gaming industry, the commercialization direction of 3D model generation applications in e-commerce and other fields is very broad.

02. From games to e-commerce, where is the "money prospect" of 3D model generation

An interesting phenomenon is that while Masterpiece X and others are still in the process of commercializing through charging points, Tripo has already focused its commercialization on platform construction and game development.

Tripo still focuses more on radiating B-end and C-end users through gaming channels.

It is understood that Tripo has established cooperation with well-known game companies, game editors and game engines at home and abroad. Its products can provide support in the early concept design, character generation, scene creation and other aspects of the game. Through testing and use, we also found that in the process of using 3D generative models to generate characters, creative inspiration can also be stimulated by adjusting the model image. Of course, saving time and technical costs is also a highlight.

One interesting issue is that Tripo created a game called TripoGo in the developer community. Players participate in the game by generating 3D models. During the duel, the winner can get the opponent's prompt to evolve the character. This game, which seems to have an old-fashioned gameplay, has the most prominent highlight of connecting C-end players and B-end game platforms together.

Based on character creation, users have a very high level of participation in the TripoGo game. While creating their own characters, they can also absorb other people's characters to evolve while enjoying the fruits of victory, which can effectively mobilize the interest of players.

Looking back at 3D generation applications such as Tripo, lowering the usage threshold is actually to provide ordinary C-end users with the opportunity to use it. Games such as TripoGo create usage scenarios for users, combining tools with scenarios, and capturing the hearts of the general C-end users. This application idea is feasible.

The application of 3D models in e-commerce is also an important commercialization direction.

Data released by e-commerce giant Amazon shows that the number of Amazon customers who viewed products through AR increased eightfold from 2018 to 2022. In 2022, Amazon statistics found that when customers clicked on AR virtual try-on or viewed product listings in 3D, the purchase conversion rate increased by an average of 2 times. 94% of customers surveyed by Amazon said they would use virtual try-on to help make purchasing decisions.

Based on this, the application of 3D generation in the e-commerce field is becoming more and more common.

Nextech3D.AI focuses its main business on the e-commerce field. The company mainly creates high-quality 3D Web AR realistic models for Amazon and other various online retailers.

Recently, Nextech3D.AI also announced that it has been certified as an Amazon Certified 3D Model External Content Provider (ECP).

Over the past five years, Nextech3D.AI has delivered nearly 100,000 3D models and millions of AR experiences to hundreds of customers, and has successfully run through the track of 3D modeling applications.

Conclusion:

The market application scenarios of 3D modeling are still promising in the future. Although there are technologically advanced companies like Tripo, more diversified commercial scenarios still need to be explored in the future. Establishing a clear technical route, absorbing more data support, and reducing the cost of technology generation are the directions that the 3D model generation track needs to continue to break through in the future.