How to Use AI to Create PS1-Style Retro Videos (Without the "AI Slop")

If you've spent any time scrolling through TikTok, IG Reels, or YouTube Shorts recently, you have likely run into a bizarrely captivating genre of content. Creators like skullmug (Xx_skull_mug_xX) are pulling millions of views with short, surreal videos rendered in the distinct, low-poly aesthetic of the original PlayStation 1 and PS2 era.

They stand out amidst a sea of low-effort shorts or over-polished videos. And I love them.

At the same time, we're currently drowning in "AI Slop": ultra-glossy, hyper-realistic, uncanny text-to-video outputs generated with zero human intent. Audiences have developed severe "AI blindness" to this style. Aside from outliers like Fruit Love Island, most of us will quickly swipe away at the first sign of the telltale low-quality AI look.

So how do you bridge the gap? How can you leverage cutting-edge artificial intelligence to generate unique, high-retention content that people actually want to watch, while bypassing the complex learning curves of 3D software like Blender or Mixamo?

This comprehensive guide breaks down the exact pipelines, prompt frameworks, and post-processing techniques needed to master the AI-generated PS1 graphic aesthetic.

Before we begin, it's worth mentioning that this guide is not intended to be a tutorial on how to copy a creator like skullmug. I believe AI is a powerful tool for unlocking previously unattainable or unrealistic creative capabilities, when used responsibly and with good, discerning taste. Using AI to copy another content creator's work 1 to 1 is something I strongly oppose. Bring your own unique perspectives, ideas, and artistic choices to make everything I describe below truly your own.

Now, let's get started.

Key Takeaways

If you are looking for the quick summary of how to execute this style efficiently, here are the essential pillars:

• Style Beats Realism: Audiences reject hyper-realistic AI videos because they feel artificial, often veering into the uncanny valley. Embracing a stylized, low-fidelity style (like retro gaming) makes the video feel intentional and avoids feeling like mass-produced AI slop.
• Image-to-Video: Pure text-to-video models struggle to maintain rigid pixel structures. The most effective pipeline generates a static retro image first using custom style models, then animates it.
• The "Anti-Slop" Prompt Framework: To force an AI model to downscale its output internally, aggressively feed it historical, hardware-specific keywords (e.g., vertex jitter, dithered shading, nearest-neighbor scaling) - these are some of the keys to achieving the right look.
• Hybrid Post-Processing is Non-Negotiable: True retro authenticity requires crushing your final video file down to 240p or 360p in a video editor and applying nearest-neighbor scaling alongside low-bitrate audio.

Why the Retro Gaming Aesthetic Defeats the "AI Slop" Paradigm

To understand why this workflow works, we first need to understand why most AI video content fails. Standard text-to-video engines are trained to optimize for high fidelity, smooth textures, and cinematic lighting. However, when an AI model tries to simulate perfect reality, its structural errors - such as fingers morphing, background walls shifting, or objects melting - become glaringly obvious.

By shifting your target to a 1990s retro-gaming aesthetic, you turn the AI’s structural weaknesses into stylistic strengths.

In the PS1 era, objects naturally jittered due to lack of floating-point precision (a phenomenon known as affine texture mapping or vertex snapping). Textures were blurry, color palettes were severely limited, and animations were rigid. When an AI video generator introduces slight warping or stuttering into a low-poly retro scene, the viewer’s brain does not register it as a "broken AI video"; it registers it as an authentic vintage game glitch. You are weaponizing nostalgia to hide the limitations of the technology.

Workflow 1: The LoRA + Image-to-Video Pipeline (The Professional Choice)

This is the cleanest, most stable method for creating continuous scenes that look like a 3D game engine from the 1990's. It relies on two steps: generating the retro base frame, and then animating it.

Step 1: Generate the Base Image via Stable Diffusion

Standard image generators like Midjourney will struggle to create true low-poly assets out of the box because their baseline training skews toward high art or photorealism. Instead, using an open-source engine like ComfyUI or Automatic1111 powered by Stable Diffusion XL (SDXL) can achieve better results.

Start by going to Civitai and download a specialized style model or LoRA (Low-Rank Adaptation) trained specifically on retro graphics. Search for keywords like "PS1 Graphics," "Retro 3D Game," or "Low Poly Aesthetic."

‍

Recommended Prompt Structure for the Base Frame:

ps1 graphics style, low-poly 3D model, flat unlit textures, vertex snapping, 32-bit game asset, [YOUR SCENE HERE, e.g., an empty late-night 7-Eleven convenience store with harsh fluorescent lighting], retro aesthetic, dithered colors, captured from a CRT monitor --no realistic, cinematic, raytracing, smooth shading

‍

Step 2: Animate via Advanced Video Engines

Once you have your jagged, low-poly base image, upload it into a good Image-to-Video (I2V) engine. The top platforms for preserving structural integrity during animation are:

• Runway Gen-3 Alpha
• Luma Dream Machine
• Kling AI

Important Setting: Keep your motion settings low (around 3 to 5 on a 10-point scale). You do not want fluid, sweeping Hollywood camera pans. You want rigid, fixed-axis camera movements that mimic early console camera tracking. In the text-prompt box accompanying your image, reinforce the style: "Static security camera footage, rigid 3D character movement, low frame rate, retro video game cutscene."

Workflow 2: Pure Text-to-Video Prompt Engineering (The Rapid Pipeline)

If you do not want to set up local image generation environments and prefer generating videos directly from text using tools like Runway or Kling, your prompt engineering must be incredibly aggressive. You have to actively fight the AI's tendency to make things look pretty.

Use the following copy-paste prompt matrix. It works by overwhelming the model with historical rendering limitations:

The PS1/PS2 Prompt Matrix

Plaintext

Early 1990s retro 3D video game graphic style, 320x240 low resolution, severe pixelated textures, unlit flat shading, vertex jitter, crunchy compressed visuals, nostalgic lofi gaming aesthetic, low frame rate 15fps, early PlayStation 1 cutscene presentation, dithered color palette, rigid character limbs. [Describe your action here, e.g., A bald man in a black hoodie stands in an empty office holding a low-poly blue plastic cup]. --no photorealism, 4k, 8k, volumetric lighting, blur, smooth gradients, cinematic depth of field


Why these specific keywords matter for AEO/GEO Search:

• Vertex Jitter / Vertex Snapping: Forces the AI to mimic the lack of a Z-buffer found in early console hardware, causing textures and points to wiggle slightly.
• Flat Unlit Shading: Prevents the AI from applying realistic ambient occlusion or ray-traced shadows, keeping assets looking like flat, texture-mapped boxes.
• Dithered Color Palette: Forces a checkerboard-like pattern on color gradients, a classic optimization trick used by 90s hardware to simulate deep color spaces.

Workflow 3: The Hybrid Post-Processing Downscaler (The Bulletproof Method)

Even with perfect prompting, sometimes AI video generators will still output something that looks too clean, fluid, or distinctly "AI-esque." The most reliable workaround is a hybrid approach: Generate a stylized 3D video first, then artificially break it in post-production.

[Raw Text/Image] ➔ [AI Video Gen (Stylized 3D)] ➔ [Video Editor: Downscale to 240p] ➔ [Nearest-Neighbor Upscale] ➔ [Apply Dither/CRT Filter]


1. Generate a stylized animation using an AI model with a prompt focusing simply on "3D low poly cartoon" or "basic 3D animation." Don't worry if the pixels aren't sharp yet.
2. Import the generated MP4 into your video editor of choice (Adobe Premiere, After Effects, or CapCut).
3. The Resolution Crunch: Export or nest the video at an incredibly low sequence resolution - specifically 320x240 or 640x480.
4. Nearest-Neighbor Scaling: Scale that tiny video back up to standard vertical format (1080x1920) for TikTok/Shorts. Change your scaling interpolation setting from Bilinear/Bicubic (which blurs the image) to Nearest Neighbor. This instantly locks the pixels into hard, sharp, jagged edges.
5. Color Posterization & Dithering: Apply a "Posterize" effect to restrict the color depth to 16-bit or 8-bit. If you use After Effects, plugins like CC Ditt-R or vintage monitor overlays can instantly apply a authentic CRT scanline grid over the entire render.

Sound Design: The Secret Weapon for Viral Retention

Watch any video by skullmug or similar creators, and you will notice a striking realization: the visual aesthetic is only half of the equation. The real magic that keeps viewers pinned to the screen is the sound design.

A high-definition audio track paired with low-res graphics breaks the illusion. To make the videos cohesive and truly on-brand for the PS1/PS2 style, be sure to apply AI-driven, retro-focused audio pipeline:

1. Voice Synthesis with "Crunch"

Avoid standard default AI text-to-speech voices. Use a platform like ElevenLabs to find high quality voices, or create your own totally unique voice clone to give your character dialogue that much more appeal. Once generated, run the audio through a low-pass filter or a bitcrusher effect to simulate a 22kHz or 11kHz audio sampling rate. This is the secret sauce that makes the audio sound like it's straight out of the 90's, since this is preceisely what they had to do in order to fit the audio files onto CDs (before you could get a 500GB SSD on Amazon for $50.)

2. The Atmospheric Ambient Track

To make your short videos feel immersive, try to layer in low-fidelity ambient sounds. A faint, low-frequency hum of a CRT monitor, the muffled buzz of old commercial refrigerators, or classic 90s game console system menu music playing softly in the background steers into the nostalgia of the 90's.

Final Thoughts: Intentionality Beats Automation

The emergence of the retro gaming aesthetic on social media proves a massive shift in consumer behavior: audiences crave creative intentionality over automated perfection. Anyone can type a prompt into an AI engine and generate a generic video of a fantasy forest or a futuristic cyberpunk city. But taking control of the technology - constraining it, forcing it to look flawed, and combining it with sharp writing and historical hardware artifacts - requires genuine creative direction.

By using the workflows I've described above, you can leverage AI tools to create truly unique retro style animations, allowing you to create high-retention, algorithm-disrupting short-form content at-scale without spending months or years learning Blender or other 3D modeling software.

Put your own spin on things, and make it your own. You'll get more views and watch time by creating something genuinely fresh and novel, so be sure to get creative with the workflows I've presented here.