Projects & Settings

Leave blank to use current origin, or point to your cloudflared tunnel.

Recent Runs

Loading runs…

Genjustsu Studio Cloudflare Pages ⚡ Super-Optimized (L40S · FP8 · TeaCache)
Wan2.2 Super-Optimized Online

Your Scene. New Characters.

The Use Case: Take any real video (dance, speech, movie scene, or meme) and replace the real person with an AI character (anime, 3D avatar, celebrity, or yourself). Super-optimized on Modal with L40S GPUs, Native FP8 Tensor Cores, and TeaCache (~1.5¢ / draft, ~4¢ / HD) while preserving the untouched original voice.

01
Source Driving Video Upload & select mode
02
Input QA Gate Review & approve
03
Character & Prompt Set references
04
Draft & 1080p HD Spawn & export
LIVE RUNTIME CONSOLE & AUDIT Waiting for activity...
[SYSTEM] Genjustsu Studio ready. Live generation logs, worker steps, and offline diagnostic traces stream here.
🎬

01 · Choose Driving Video & Mode

Upload a performance clip (up to 30s). The original soundtrack will be preserved intact.

🎬 Direct RGB (Native) Zero Token

Direct driving video latents straight into Wan2.2 with FP8 Tensor Cores & TeaCache. Super-optimized to ~1.5¢ per draft.

🎯 Best for: Full-body motion, action, dance, running & everyday videos.
🗣️ Fast Face Mesh Lipsync Pro

Tracks 468 digital facial motion-capture landmark dots without using slow depth inference.

🎯 Best for: Talking-head dialogue, podcasts, close-up interviews & speeches.
🎭 Depth + SAM 3D Isolation

Generates 3D depth geometry + raw SAM 3 cutout masks to isolate people from complex backgrounds.

🎯 Best for: Moving cameras & crowded backgrounds with furniture or objects.
👑 Depth + Face Mesh Cinematic

The ultimate combo: extracts 3D background depth while tracking high-fidelity facial performance.

🎯 Best for: High-end cinematic acting scenes with speech inside busy 3D sets.
💡 What is this app & which mode should you pick?

Simple Explanation: You upload a video of a real person acting, talking, or dancing, plus a reference photo of an AI character (anime, 3D character, digital avatar, or yourself). The AI swaps the character into the scene while preserving the exact body timing, camera angle, facial expressions, and original voice!

🎬 Direct RGB Use for dancing, sports, action, or whole-body videos. Zero external tokens needed.
🗣️ Fast Face Mesh Use when someone is talking closely to the camera and exact mouth and lip-sync matter most.
🎭 Depth + SAM Use when the background is messy or moving so the character doesn't blend into the scenery.
👑 Depth + Face Mesh Use for cinema-grade movie clips needing both 3D background separation and fine lip movement.
🔍

02 · Input Review Studio & Cryptographic Gate

Inspect the full clip. Generation is strictly blocked until approval is signed to the SHA-256 fingerprint.

Review Required
Prepared video preview Download Clip ↓
Fingerprint: None
🎭

03 · Character Identity & Creative Prompt

Supply your target character reference image. Modal Wan2.2 will transfer appearance seamlessly.

Modal H100 Ready
+ Photorealistic + 8K Masterpiece + Strict Lip-Sync + Anime Style
🌟

04 · Draft Preview & Approved HD Export

Draft generated at 480p. When satisfied, trigger 1080p full-quality rendering on Modal.

Draft Ready
Preview with 100% original untouched source audio restored Download Result ↓

Satisfied with the motion and lip-sync?

Approve this draft to launch HD rendering on Modal serverless H100.