Every asset, from a prompt.

Generates reference images, concept art and icons with multi-model fan-out across fal.ai and OpenAI, cuts subjects out onto transparent backgrounds, turns text or images into static props and auto-rigged, animated 3D characters through Meshy or fal — then remeshes, UV-unwraps or re-textures them on request — produces video from text or images with a matching soundtrack, and drops the results straight into Unreal as textures, meshes and more. No manual import step.

Pulls ready-made assets straight into your project: your owned Fab and Megascans library (with Epic login and automatic fallback to an older engine build when a listing has no exact match), Sketchfab models, CC0 models, textures and HDRIs from Poly Haven and ambientCG, Quaternius low-poly packs, VRoid Hub avatars, or any direct URL, ZIP or local folder. Dry runs show what will land where before anything is written, and every import is tagged with its source and license.

Drives Blender as a separate process straight from the editor — no add-on to install. Cleans and repairs meshes, remeshes and decimates, unwraps UVs, bakes high-to-low normal and AO maps, auto-rigs with Rigify, converts .blend/.obj/.usd files and splits kitbash scenes, then imports the result as a real Unreal asset. Round-trips UE assets through Blender and back, runs bpy code, and speaks the mcp-for-blender feature set — scene info, viewport screenshots, Poly Haven, Sketchfab and AI 3D generators.

Generates sound effects, NPC dialogue voiceover (including voice cloning and voice-library management) and full music tracks from text prompts via ElevenLabs — imported directly as USoundWave assets, ready to plug straight into Sound Cues and MetaSounds.

The newest skill in the set generates whole coherent families of web and presentation media — landing-page hero videos, card grids, slide backgrounds, OG/social cards, icon and spot-art sets — from a single LLM prompt. It measures the target format first (aspect, pixel size, crop behaviour, text safe-zones), reuses one style spec across the whole set so nothing looks like it came from five different designers, fans out across multiple image and video models with a per-slot pick, and lets ffmpeg do the finishing work to the exact spec your CSS or slide master expects — delivered as files on disk, with audio covered by its sibling skill.