If you've been on the internet in the past few years you’ve likely seen the viral AI historical figures videos popping up all around social media. And if you're curious on how to recreate these cool videos with Artificial intelligence historical figures, you're in the right place. With 

, it's easier than ever to create your very own AI generated historical figures videos and bring history to life in a whole new way using 

In this blog, we'll explore how all this works and showcase three examples of historical figures brought to life through AI.

Here’s how it works: First, we upload an image or multiple images of historical figures into Kling API or any AI video generation model of your choice. Next, we add a prompt specifying what we want the historical figures to do in the video. And finally, we just wait for Kling API or the AI video generation model to generate the video for you!

For our first example, we’ll focus on the most recognizable pharaoh, Cleopatra. We’ll be using the famous painting "Cleopatra" by John William Waterhouse, created in 1888. This masterpiece is currently housed in the Tate Britain Collection in London.

Next, the image will be put into Kling API, along with the prompt shown below the GIF. We find it really impressive how fluid and smooth the motions are. We’re especially drawn to the fingers of the right hand, which point directly at the viewer. Unlike many AI-generated visuals, where fingers can often be a weakness, in the video of Cleopatra generated by AI we made, they are perfectly generated and animated.

For our second example, we’ll be using this iconic portrait of George Washington, which serves as his presidential portrait. Painted by Gilbert Stuart in 1803, this renowned portrait is titled simply "George Washington." It is now housed in the National Gallery of Art collection in Washington, D.C.

Next, we uploaded the image of George Washington into Kling API with the prompt shown below. We believe the animation is both smooth and realistic, particularly the way George Washington smiles and laughs. The subtle wrinkles around his eyes as he smiles add a touch of realism.

For our final example, we are going to do something a little bit different. nstead of a painting, we’ll be using a photograph of the statue Alexander the Great, signed by Means of Pergamon, son of Aias, who is believed to be the sculptor. Created in the 3rd century BCE, this statue is now housed in the Istanbul Archaeology Museum.

Next, we uploaded the image into the Kling API along with the prompt, which you can see below the video. The vide looks pretty cool and aesthetically pleasing. The fingers are generated without any issues along with the shadow that the hand casts.

As you can see from the three examples above AI is quite good at bringing historical figures to life, especially the Alexander the Great and Cleopatra videos. Although this is not an AI reconstruction of the historical figure (We might do this in a later blog!), AI image to video generation has come a long way.

We're excited to see how people will keep using AI in even more fun and creative ways, just like this!

In this blog we use Kling API to bring historical figures to life!

Bring Historical Figures to life using AI like Kling API

TRELLIS is a large-scale 3D asset generation model open-sourced by 

 that supports high-quality 3D content generation from text or images. It employs a structured 3D latent space approach to achieve scalable and versatile 3D generation.Key Features:

• Supports both image-to-3D and text-to-3D generation(soon) modes

• Uses structured 3D latent space approach for higher generation quality

• Provides multiple 3D representation formats (Gaussian point clouds, radiance          fields, meshes, etc. 

• Supports export to standard 3D file formats like GLB/PLY

The Trellis API is offered by PiAPI based on the Trellis model. Given their accelerated hardware infrastructure and custom-developed inference framework, PiAPI is able to provide APIs for these two models at very market competitive costs while maintaining performance and minimizing latency.

To understand the practical differences and capabilities of TRELLIS, let’s compare it with another AI model , 

 by exploring how each generates GIFs of knee-high Converse shoes. This example will illustrate not only the visual quality of outputs but also highlight the efficiency and versatility of TRELLIS’s structured 3D latent space compared to its counterpart.

While Meshy AI’s model attempts to replicate the form of a sneaker, it      lacks refinement, with rough edges and inconsistent proportions that detract from its overall realism. The design feels unfinished, and the exaggerated textures give it a somewhat artificial appearance.

Trellis3D offers a more structured and defined design, with a consistent and cohesive shape. While it isn’t perfect, the overall form aligns more closely with the natural proportions and flow of a real sneaker, making it slightly more believable in terms of structure.

In Meshy AI, the textures are overly dramatic, with exaggerated shading and surface marks that can make the model feel less cohesive. The material representation lacks subtlety, and the surface details do not align naturally with real-life fabric or rubber.

The textures in Trellis3D are simpler but more balanced. While not highly detailed, the model avoids the harsh highlights and oversaturation seen in Meshy AI, giving it a more understated and realistic appearance.

In Meshy AI, the elements of the model, such as the laces, sole, and fabric, feel disjointed. The lack of polish in its construction results in a model that appears incomplete and inconsistent.

Trellis3D delivers a more cohesive presentation. Although less intricate, the elements are better aligned and feel more harmonious, providing a sense of completeness despite the lack of fine detailing.


In summary , while Meshy AI showcases an ambitious design with intricate textures, its lack of polish and cohesion limits its effectiveness. Trellis3D, although simpler, provides a more balanced and structured model, making it the better choice for realism and overall presentation in this comparison.
While the first comparison highlighted TRELLIS's ability to deliver realistic textures and details for modern objects like shoes, its strengths become even more apparent when applied to historical or thematic objects. Let’s now compare TRELLIS and Meshy in generating a medieval weapon model to further illustrate its capability to balance ruggedness, detail, and realism.

In Trellis 3D , the model features rugged textures and a worn appearance, which gives it an aged and historical look. However, the form appears less refined, with a somewhat irregular shape that may detract from its believability as a weapon.

In Meshy AI, the model has a sleek and well-defined structure, with smooth yet balanced textures that reflect modern design standards while still retaining a worn, medieval aesthetic.

In Trellis3D, textures are rugged and emphasize wear and tear but appear overly simplistic and lack finer details such as grain or specific surface imperfections.

In Meshy AI, the textures are more detailed, with shading and subtle surface marks that provide depth and a realistic appearance. The material representation (wood/metal) is also more recognizable in Meshy AI.

In Trellis 3D, while the aged texture aligns with medieval themes, the model lacks structural precision and detail, which may compromise its overall believability.

In Meshy AI, the model is more cohesive and complete , with proportional design and polished finishing, making it more convincing as a weapon.

In summary, Meshy AI is the better model overall due to its superior structural refinement, detailed textures, and balanced realism. While Trellis3D captures ruggedness, it lacks the precision and detail required to compete with the more polished and visually engaging Meshy AI model.


Building on the comparison between Trellis3D and Meshy AI, a similar focus on texture detail, realism, and structural refinement can be seen when evaluating Slitherwing Dragon Egg using Trellis3D and another image-to-3D Model,Tripo3D.

The layering and arrangement of the petals in Trellis are more cohesive and natural, closely mimicking the input image .

Each petal appears distinct, with clear thickness and spacing, showing meticulous design work that conveys the model's completeness.

 texture appears smoother, with fewer harsh highlights or reflections. This provides a more refined and polished look compared to Trellis3D.

Tripo3D displays better control over lighting and surface reflection, preventing oversaturated bright spots. This makes the texture more visually appealing and realistic.

In Tripo3D, the color distribution seems more uniform and less patchy, enhancing the object's overall natural appearance.

Tripo3D’s textures have a subtle gloss that gives the object a more organic, life-like feel, compared to the overly reflective or shiny look in Trellis3D.

In summary, Trellis demonstrates greater attention to sharpness and high-contrast design, whereas Tripo3D provides better realism and natural texture.

Jean Claude Van Damme from bloodsport with red shorts

In Trellis 3D, the model features sharper and more defined contours, with clear outlines and visible muscle structure. However, the lighting is harsher, making the model appear slightly less natural and organic.

In Tripo3D, the model has a smoother and softer design, which gives it a more cohesive appearance. However, the lack of detail in textures and overly smooth surfaces detracts from its realism and believability.

In Trellis 3D, textures are well-defined and exhibit noticeable contrast, particularly in the highlights and shadows, enhancing the model's features. However, the highlights may appear exaggerated, making the surface feel less lifelike.

In Tripo3D, textures are more uniform and subdued, lacking the intricate highlights and shadows necessary for dynamic depth. This results in a flatter appearance and reduces the overall visual impact.

In Trellis 3D, the model demonstrates better definition and structure, with visible attention to proportion and details. However, the lighting contrast might feel overly sharp in certain areas, affecting its cohesion.

In Tripo3D, the smoother and less detailed design gives a cohesive but overly simplified look. The absence of distinct textures and fine details impacts the model’s completeness and realism.


In summary, Trellis3D is the better model due to its sharper design, clear textures, and better definition, which make it more visually appealing and believable. While Tripo3D offers a smoother and more cohesive appearance, it lacks the detail and realism needed to compete with the refined and dynamic Trellis3D model.

The comparisons throughout this analysis demonstrate the nuanced strengths and weaknesses of Trellis3D, Meshy AI, and Tripo3D in 3D model generation. Trellis3D stands out in many cases for its sharpness, structural clarity, and ability to deliver realistic, detailed models, making it a reliable option for producing polished assets. While Meshy AI and Tripo3D excel in smoothness, lighting balance, and modern aesthetics, their lack of refinement and cohesion in certain models highlights room for improvement.

Trellis3D’s capability to achieve rugged realism, proportional accuracy, and cohesive design across different scenarios—from shoes to medieval weapons—showcases its versatility and adaptability for diverse applications. This combination of detailed textures and structural completeness gives Trellis3D an edge over its competitors in specific use cases.

Ultimately, the choice of model depends on the priorities of the user—whether precision, realism, or stylistic elements are paramount. However, the examples explored in this blog underline Trellis3D’s capacity to deliver consistent and high-quality outputs, making it a compelling option in the realm of 3D content generation.

And if you are interested in any of the other 

PiAPI has recently launched our very own 3D Trellis API!

3D Trellis API through PiAPI

Hi developers! We are very delighted to make this tutorial to show how you can use 

Before we dive into this tutorial, do sign up for our Workspace, obtain your API Key, and check out the documentation for our Faceswap API endpoints.


Firstly, you will need to open up or download Postman. Click the "New" button on the top left of your screen and create a new HTTP request, then change the request type to "POST".

Enter the Faceswap API's endpoint address from our docs into the endpoint input field, click on "Headers" and type "x-api-key" in the empty space under the "Key" text and enter your API Key under the "Value" text as shown in the example below.

Next, click on "Body" located below the endpoint address, then click on "raw", copy the code from our docs and insert it into the body section. Press "Send", then a task id will be returned to you in the response section, and you will need to keep it for later use. Once you're done, proceed to the next step.

Now, we are going to fetch for the task that has been submitted. You will need to create a new HTTP request, change the request type to "POST", then copy the fetch endpoint URL from our docs into the endpoint input field, similar to the above steps.

Then, enter the saved task id response into the body of the fetch endpoint, and press "Send". Once the result status shows "finished", you may copy the returned URL (as per arrow below) into a web browser and see your face-swapped image!

Note: As you are sending the fetch request, if the returned "status" is either "pending" or "processing", then you should wait for a moment before you re-send the request again.

If you are wondering about pricing, please check out our 

's pricing section for more information. We also provide custom dedicated deployment service to users who need solutions for specific requirements (ex. low latency, high concurrency, shorter queuing time).

And this is how you can get started using PiAPI's Faceswap API with Postman! Thank you for reading through this tutorial. Do share with people who find this helpful and don't be shy to contact us with any feedback you might have!

Interested in using Faceswap API but don't know how to do it? Check out this tutorial on how to use Faceswap API with Postman and create your favourite face swapped images!

How to use Faceswap API from PiAPI with Postman

Tutorial

Conclusion

More Stories

Bring Historical Figures to life using AI like Kling API

3D Trellis API through PiAPI