Let's Create! — On-Device & WebGPU Prompt Studio
Hugging Face CDN 스트리밍 로더 및 AMEVA-Forge WebGPU 텐서 가속 엔진을 통해 브라우저 클라이언트에서 직접 생성되는 애니메 디퓨전 스튜디오입니다.
[쇼룸 프리뷰 안내 / Interactive Studio Notice]
본 웹 스튜디오는 인터랙티브 파라미터 튜닝 및 사전 컴파일된 레퍼런스 결과 시각화를 제공하는 쇼룸입니다. 가상화(PRoot) 없는 순수 100% 온디바이스 네이티브 하드웨어 연산(Cortex-A78 x4 61초 실측)은 Termux CLI (pip install termux-diffusion)에서 독립적으로 구동됩니다.
MASTER PROMPT SPECIFICATION
54 CLIP Tokens
Negative:
low quality, blurry, extra limbs, watermark, text
WebGPU Ready · Animagine XL · Latency: 1.84s
Client-side Real Generative Pipeline · Zero Server Egress
Native Code Execution (On-Device & WebGPU SDK)
선택한 프롬프트 및 모델 파라미터가 AMEVA-Forge WebGPU 및 Termux SDK 코드로 실시간 동기화됩니다.
import { ForgeDiffusion } from "@uno-km/ameva-forge";
// 1. Initialize WebGPU Denoising Pipeline with animagine-turbo CDN
const pipeline = await ForgeDiffusion.fromPretrained("cagliostrolab/animagine-xl-3.1", {
device: "webgpu",
precision: "fp16"
});
// 2. Execute zero-server 4-step diffusion generation
const image = await pipeline.generate({
prompt: "cute orange cat surfing on a small wave, centered composition, cat clearly visible in the middle of the image, wet reflective ocean surface, playful expression, fluffy fur slightly wet, bright blue sky above, soft white clouds, sunlight reflections on water, dynamic splash effects, cinematic lighting, highly detailed, vibrant colors, adorable, clean composition, focus on the cat, mobile wallpaper, masterpiece, no text, no watermark",
steps: 4,
cfg: 1.5,
seed: 42891,
width: 512,
height: 512
});
image.renderToCanvas(document.getElementById("outputCanvas"));