Increase REALISM of 3D images with Magnific AI

Nuno Silva
8 May 202409:08

TLDRIn this tutorial, the host collaborates with Vasilis from VW Art Club to demonstrate how AI post-production can transform 3D renders into near-photorealistic images using Magnific AI. The tool features two main functions: 'upscale' for enhancing image resolution and adding AI-driven creativity, and 'style transfer' for applying styles from one image to another. The process involves several iterations with different settings and is not instant, requiring manual tweaking to achieve the best results. The video showcases the step-by-step evolution of a render, starting from a basic clay render to a hyper-realistic image, with attention to details like vegetation, reflections, and textures. The host emphasizes the importance of experimenting with AI settings and combining them with manual editing in Photoshop for the final touch. The result is a significantly improved visual appeal, showcasing the potential of AI in enhancing architectural and design visuals.

Takeaways

  • 🎨 **Using Magnific AI**: The tutorial demonstrates the use of Magnific AI for enhancing 3D renders with features like upscale and style transfer.
  • 📈 **Upscale Feature**: Magnific AI's upscale feature not only increases image resolution but also allows for AI-driven creativity adjustments.
  • 🖼️ **Style Transfer**: The style transfer feature applies the style of one image to another, creating a unique and realistic look.
  • 🔍 **Detail Improvement**: The process involves several iterations with different settings to refine details and textures in the image.
  • 💡 **Creativity Slider**: Adjusting the creativity level lets the AI add more details but can also lead to unexpected results if set too high.
  • 🌟 **HDR and Resemblance**: The HDR setting enhances image definition, while resemblance keeps the generation close to the original image.
  • 📝 **Prompts for Guidance**: Using detailed prompts helps guide the AI to achieve the desired level of realism and style.
  • 🧩 **Photoshop Integration**: The final image is often a composite created by merging different AI-generated layers in Photoshop.
  • 🔧 **Iterative Process**: It's important to experiment with various settings and generations to achieve the best results.
  • 🏡 **Architectural Focus**: The example focuses on transforming a simple clay render of a house into a hyper-realistic image.
  • ⏱️ **Time Investment**: The process can be time-consuming, involving several generations and style transfer adjustments, but yields significant improvements.

Q & A

  • What is the main purpose of Magnific AI as described in the tutorial?

    -The main purpose of Magnific AI is to enhance the realism of 3D images through AI-driven post-production techniques, including upscaling and style transfer.

  • Who collaborated with the presenter for this tutorial?

    -The presenter teamed up with Vasilis from VW Art Club, a popular Facebook group for 3D artists and architects.

  • What are the two main features of Magnific AI highlighted in the tutorial?

    -The two main features are upscaling and style transfer.

  • How does the upscaling feature benefit the images?

    -The upscaling feature boosts the resolution of images and allows for various adjustments to add AI-driven creativity, enhancing the overall visual appeal.

  • What does the style transfer feature do?

    -The style transfer feature allows users to apply the style of any image to their own, transforming the visual style and realism of their renders.

  • How many tries does it usually take to get the desired result with AI adjustments?

    -It usually takes about five or six tries, each with different settings, to achieve the desired result.

  • What should be the maximum resolution for the input image during the upscaling process?

    -The input image should have a maximum resolution of 2K to avoid wasting credits on the scale factor.

  • What was the initial setting for the 'creativity' slider in the tutorial example?

    -The initial setting for the 'creativity' slider was zero.

  • What happens when the 'creativity' value is set too high?

    -Setting the 'creativity' value too high can lead to strange results, as the AI may add unexpected details that deviate significantly from the original image.

  • What was the impact of the style transfer in the final generation of the render?

    -The style transfer significantly enhanced and refined the final image, adding more details and textures that were not as pronounced in the original render.

Outlines

00:00

🎨 AI Post-Production for Realistic Renders

This tutorial introduces viewers to the process of enhancing renderings using AI post-production. The video is a collaboration with Vasilis from the VW Art Club, focusing on a selected project submitted by a member. The tool of choice is Magnific AI, which offers two main features: 'upscale' for improving image resolution and 'style transfer' for applying the style of one image to another. The presenter, Noo, explains that while the tool is powerful, it requires some manual tweaking and multiple iterations to achieve the best results. The process involves loading the raw render into Magnific AI, adjusting settings like creativity, HDR, resemblance, and engine, and experimenting with different generations. The goal is to add realism and detail to the image, which is then further refined in Photoshop by merging different layers and making final adjustments.

05:01

📈 Enhancing Details with AI and Style Transfer

The second paragraph delves deeper into the AI post-production process, highlighting the use of a detailed prompt and the importance of refining specific parts of the image, such as glass and foreground vegetation. The presenter demonstrates how increasing the creativity level in AI can lead to unexpected and sometimes wild results, like adding houses in trees, but also how useful elements like improved leaves and water reflections can be extracted for use. The focus then shifts to refining the image by adjusting settings to enhance details and textures. The use of style transfer is introduced as a way to apply a different render's style to the image, resulting in a significant transformation from the original. The final step involves fine-tuning the image in Photoshop, incorporating elements like water pebbles, vegetation, and a metal sheet on the roof to add depth and polish the final look. The presenter emphasizes the value of time invested in this process, which can significantly improve the final image's visual appeal.

Mindmap

Keywords

💡Magnific AI

Magnific AI is a tool used to enhance the realism of 3D renders through AI post-production. It offers features such as upscale and style transfer, which are used to boost image resolution and apply the style of one image to another, respectively. In the video, it is demonstrated how this tool can transform a basic render into a highly realistic image.

💡Upscale

Upscale is a feature in Magnific AI that increases the resolution of images, enhancing their detail and clarity. The video shows how this feature can be used to improve the visual quality of raw renders, making them appear more realistic.

💡Style Transfer

Style Transfer is a feature in Magnific AI that allows users to apply the style of one image to another. This is used to add creative elements and enhance the overall look of the render. The video explains how style transfer can be used to pull out details and textures, making the final image more appealing.

💡Creativity Slider

The Creativity Slider in Magnific AI controls the level of AI-driven creativity applied to an image. Higher values allow the AI to add more details and deviate from the original image, while lower values keep the image closer to the original. The video illustrates the effects of adjusting this slider through various generations of the render.

💡HDR

HDR (High Dynamic Range) is a setting in Magnific AI that enhances the definition and detail of images. In the video, different HDR values are tested to show how they affect the final image, with higher values increasing detail but potentially introducing artifacts.

💡Resemblance

Resemblance is a property in Magnific AI that determines how closely the generated image resembles the original input. Lower values give the AI more freedom to be creative, while higher values keep the image closer to the original. The video shows how tweaking this setting can affect the final render.

💡Photoshop

Photoshop is a graphics editing software used to merge different generations of AI-enhanced images. The video demonstrates how Photoshop is used to combine elements from various iterations to achieve the best final result.

💡Generations

Generations refer to the different iterations of image processing using Magnific AI. Each generation applies various settings to the render, progressively enhancing its realism. The video details how multiple generations are used to refine and improve the final image.

💡AI Post-Production

AI Post-Production involves using artificial intelligence tools to enhance and refine images after the initial rendering process. The video showcases how Magnific AI is used in post-production to transform basic renders into highly detailed and realistic images.

💡VW Art Club

VW Art Club is a popular Facebook group for 3D artists and architects. The video mentions a collaboration with Vasilis from VW Art Club, highlighting the community's involvement in selecting projects for AI enhancement using Magnific AI.

Highlights

This quick tutorial will show you how your renders can go from this to an almost real photo with AI post-production.

I've teamed up with Vasilis from VW Art Club, a popular Facebook group for 3D artists and architects.

We are going to use Magnific AI, which has two features: upscale and style transfer.

The upscale feature boosts the resolution of your images and adds AI-driven creativity to make your image pop.

The style transfer feature allows you to apply the style of any image to your own.

For example, Jav Lopez transformed a simple clay render into a stunning realistic image using Magnific AI.

You need to experiment with different settings to achieve the best results with AI.

It usually takes about five or six tries, each with different settings, and then you merge them all together in Photoshop.

For the first generation, we used the upscale feature with a raw render at a maximum resolution of 2K.

Higher HDR values increase the definition and detail but can create artifacts.

Increasing the resemblance makes the generation closer to the original image, but very high values can give some artifacts.

In the second generation, we used a prompt that said 'a hyper realistic rendering of a modern house in a lush green forest'.

In the next round of edits, we bumped up the creativity level to five, which gives the AI more freedom to be creative.

In the following generation, the creativity was cranked up to eight, allowing the AI to go wild, adding houses in the trees.

In the latest generation, we used style transfer, starting with the raw render as a reference and the clay version as the input image.

The goal is to use these tools to enhance and refine the final result, pulling out details and textures that weren't as pronounced before.

The final result from the style transfer shows significant improvements in the foreground vegetation and the metal sheet on the roof.

The process might take up to an hour, but it can significantly improve the final image by layering techniques to enhance and refine the visuals.

The texture repetition was kept on purpose to maintain a subtle 3D feel.