AI Image Generation Setup

Let AI seamlessly extend your wallpapers to fit any screen.

What Is AI Border Fill?

When a wallpaper’s aspect ratio doesn’t match your monitor (for example, a 4:3 photo on a 16:9 screen), BioniX traditionally stretches, crops, or adds black bars. With AI Border Fill, BioniX uses artificial intelligence to extend the image borders naturally — generating new content that blends seamlessly with the original.

The result is cached locally so each wallpaper is only processed once. You can generate on demand (one image at a time) or pre-process your entire playlist in the background.

Local vs. Cloud Providers

BioniX supports two categories of AI providers. Pick whichever suits you best.

Local Providers

AI runs on your own computer using your GPU. Your images never leave your machine.

  • Completely free (no API fees)
  • Full privacy — nothing is uploaded
  • Requires a dedicated GPU (NVIDIA recommended)
  • Slower (30 sec–4 min per image depending on hardware)

Cloud Provider (Stability AI)

AI runs on remote servers. You send images via the internet and receive results back.

  • No GPU required — works on any PC
  • Fast (5–30 seconds per image)
  • Pay per image
  • Images are sent to third-party servers

Setting Up ComfyUI

Local Free Most flexible option — supports any model you install.

  1. Download ComfyUI from github.com/Comfy-Org/ComfyUI and follow its installation instructions.
  2. Install an inpainting model. Download an inpainting checkpoint (for example, sd-v1-5-inpainting.ckpt or a Stable Diffusion XL inpainting model) and place it in ComfyUI’s models/checkpoints/ folder.
  3. Start ComfyUI. It usually runs on http://127.0.0.1:8188 by default.
  4. In BioniX Settings → AI tab, select ComfyUI as the provider and set the server URL to http://127.0.0.1:8188 (the default).
  5. Click “Test Connection”. BioniX will connect to ComfyUI and verify that an inpainting model is available. A green checkmark confirms everything is ready.
Tip: BioniX ships with a built-in outpainting workflow that uses only standard ComfyUI nodes — no custom nodes required.

Setting Up SD WebUI / Forge

Local Free Simpler local alternative. Forge is the actively maintained fork with better performance.

  1. Download Forge from github.com/lllyasviel/stable-diffusion-webui-forge (or the original Automatic1111 WebUI).
  2. Launch with the API flag. Edit the startup script and add --api to the command-line arguments. For example: webui.bat --api
  3. Install an inpainting model if you don’t have one already. Place it in the models/Stable-diffusion/ folder.
  4. In BioniX Settings → AI tab, select SD WebUI / Forge and set the URL to http://127.0.0.1:7860 (the default).
  5. Click “Test Connection” to verify.
Important: The --api flag is required. Without it, the WebUI won’t accept requests from BioniX. If you forget it, the “Test Connection” button will report a connection error.

Setting Up the Cloud Provider

The cloud provider, Stability AI, needs an API key. Sign up on its website, generate a key, and paste it into BioniX.

Stability AI

Cloud Dedicated outpainting endpoint.

  1. Create an account at platform.stability.ai
  2. Go to API Keys and generate a new key
  3. In BioniX, select Stability AI as provider and paste the API key

Cost Comparison

Estimated cost to process a typical playlist of 500 wallpapers (assuming ~30% need border fill = 150 images).

Prices are estimates; see the provider’s pricing page.

Provider Per Image 150 Images Speed
ComfyUI / Forge (local)FreeFree30s – 4 min
Stability AI~$0.04~$65 – 15s

Settings Explained

Enable AI border fillMaster switch. When off, BioniX uses traditional resize modes.
ProviderChoose which AI service to use for image generation.
Server URLFor local providers only. The address where ComfyUI or SD WebUI/Forge is running.
Test ConnectionVerifies the provider is reachable. For ComfyUI, also checks that an inpainting model is installed.
API KeyFor cloud providers only. Your authentication key from the provider’s website.
Quality sliderBalances speed and quality. “Fast” uses fewer AI steps (faster, slightly lower quality). “Best” uses more steps (slower, higher quality).
PromptInstructions for the AI. The default works well for most images. Customize if you want a specific style.
Negative promptThings to avoid in the generated areas (e.g., “ugly, distorted, watermark”).
Processing modeOn demand: generates when each wallpaper is displayed. Background: pre-processes your entire playlist ahead of time.
CacheAI-generated images are cached locally so each wallpaper is only processed once. You can clear the cache at any time to free disk space.

GPU Recommendations for Local Providers

Local AI image generation is GPU-intensive. An NVIDIA GPU with at least 6 GB of VRAM is recommended. AMD GPUs work with some setups but NVIDIA has the broadest compatibility.

GPUVRAMApprox. Speed (1024×1024)
RTX 306012 GB~3–4 minutes
RTX 407012 GB~1–2 minutes
RTX 409024 GB~20–40 seconds
RTX 508016 GB~30–90 seconds

Don’t have a powerful GPU? No problem. The cloud provider works on any computer — even without a dedicated graphics card.

Frequently Asked Questions

Do I need an internet connection?

Only for cloud providers. Local providers (ComfyUI, SD WebUI/Forge) run entirely on your machine with no internet required.

Are my wallpapers uploaded anywhere?

With local providers: no. Your images stay on your computer. With cloud providers: images are sent to the provider’s servers for processing. If you do not want that, pick a local provider.

What happens if the AI provider is unavailable?

BioniX falls back to its standard resize mode (stretch, crop, or letterbox). No error is shown — it just works the traditional way.

Can I use this with multiple monitors of different sizes?

Yes. BioniX generates a separate variant for each unique monitor resolution. If you have a 1920×1080 and a 2560×1440 monitor, both get their own optimized version.

What about 4K or ultrawide monitors?

Most AI models have a maximum output resolution. When the AI returns a smaller image, BioniX upscales it to your monitor’s resolution. The result is cached at full resolution.

How much disk space does the cache use?

Each cached image is typically 2–8 MB. A playlist of 500 images, of which about 150 need border fill, would use roughly 300 MB–1.2 GB of cache. You can see the current cache size and clear it from the AI settings tab.

The “Test Connection” fails. What should I check?

  • Make sure the AI provider is running (ComfyUI or SD WebUI/Forge must be started separately)
  • Verify the server URL and port number in BioniX settings
  • For SD WebUI/Forge, make sure you started it with the --api flag
  • Check that your firewall isn’t blocking the connection
  • For cloud providers, verify your API key is correct and your account has credit

Need More Help?

Can’t get it working? Send us a message and we’ll help you set it up.

Contact Us

Like BioniX? Get It for Free.

After the 15-day trial, keep BioniX free through the community program: post about BioniX, and how far your post reaches decides your membership.