Skip to main content
Unlisted page
This page is unlisted. Search engines will not index it, and only users having a direct link can access it.

Stable Diffusion XL

The Stable Diffusion XL (SDXL) integration in ACT3 AI enables high-quality image generation for set backgrounds, concept art, and character reference imagery. SDXL is an image model rather than a video model — it produces static visuals that you use for planning, previsualization, and as inputs to video generation tools like Flux.

What Stable Diffusion XL Does​

SDXL generates high-resolution, photorealistic or stylized still images from text descriptions. You can use the outputs as:

  • Set backgrounds for scene environments
  • Character concept art to guide digital actor creation
  • Texture references for Blender 3D modeling
  • Pitch deck imagery for presentations
  • Visual planning references before committing to video renders

Key Capabilities​

  • High-Resolution Image Generation — Up to 2048×2048 pixels for detailed visual concepts
  • Prompt-Based Styling — Describe the look, mood, and subject; SDXL generates matching visuals
  • Style Consistency — Apply custom models or LoRA files to maintain a consistent aesthetic
  • Image-to-Image Mode — Transform existing images into variations while preserving composition
  • Batch Output — Generate multiple variations from a single prompt for faster visual development

How to Use​

  1. In the Editor, click AI → Stable Diffusion XL
  2. Enter your prompt, including style, setting, lighting, and subject details
  3. Optionally upload a base image for Image-to-Image transformation
  4. Adjust resolution, seed, and model settings as needed
  5. Click Generate and review results in the Asset Library
  6. Use accepted images as set backgrounds or concept references

Image-to-Image Mode​

Image-to-Image mode transforms an existing reference image into a new variation:

  • Upload a sketch, reference photo, or previous generation
  • Describe what the new version should look like
  • SDXL preserves the composition while applying the new style and content description
  • Use this to iterate on a visual concept without starting from scratch

Credit Usage​

SDXL image generation uses credits based on resolution and number of variations:

  • 1024×1024 image: approximately 0.5 credits
  • 2048×2048 image: approximately 1 credit
  • Image-to-Image costs the same as Text-to-Image at equivalent resolution
  • Batch generation (multiple variations) costs per image generated

Best Use Cases​

  • Visual Planning — Generate reference images quickly before committing to video renders
  • Set Design — Create background imagery for fantasy, sci-fi, or period environments
  • Character Concepts — Visualize character appearance options before building digital actors
  • Blender Textures — Generate texture images for 3D surface materials
  • Animatics — Use SDXL images as the visual input for Flux's Image-to-Video workflow

Combining SDXL with Other Tools​

SDXL works particularly well in combination with other ACT3 AI features:

  • Generate concept images with SDXL, then animate them with Flux Image-to-Video
  • Create set backgrounds with SDXL, then import to Blender for 3D use as environment textures
  • Build character concept art with SDXL, then use it as a reference when creating Digital Actors
  • Run complex SDXL pipelines through ComfyUI for multi-pass compositing

Prompt Tips​

  • Be descriptive — include camera style, lighting conditions, color palette, and mood
  • Use consistent keywords across multiple prompts to maintain visual coherence within a project
  • For scene matching, start with an Image-to-Image workflow using a reference frame from your existing shots
  • Save prompts that produce good results for reuse across similar shots

Best Practices​

  • Use SDXL early in production for cheap visual exploration before committing to video renders
  • Generate 3–5 variations per key shot and choose the best as a visual reference
  • Keep consistent style prompts across related shots for visual coherence
  • Save accepted images to the Asset Library with descriptive tags for easy retrieval

Troubleshooting​

Images don't match the style you described — Be more explicit about art style keywords (e.g., "cinematic photography," "oil painting," "digital art," "photorealistic"). Vague style descriptions produce inconsistent results.

Composition is wrong — Use Image-to-Image mode with a rough sketch or reference image to constrain the composition.

Batch generation consuming too many credits — Reduce the number of variations per prompt and select only what you need.