Votre visage. Votre stream.

I am setting up the FaceNinja Creator Kit in this folder. I do not write code, so explain each step in one short
sentence and ask me before anything that needs my password, installs a driver, or changes system settings.

1. Tell me my operating system, CPU, RAM, graphics card and free disk space, and compare them with the
   "What you need" table in GUIDE.md. Warn me if something is below the minimum.
2. Check what is installed and tell me what is missing:
   - Unreal Engine 5.8 (Windows: C:\Program Files\Epic Games\UE_5.8; Mac: /Users/Shared/Epic Games/UE_5.8).
     If it is missing, do not install it yourself: tell me to install Unreal Engine 5.8 from the Epic Games
     Launcher's Library tab (default options), and wait until I say it is done.
   - OBS Studio. Offer to install it (Windows: winget install OBSProject.OBSStudio; Mac: brew install --cask obs).
   - A virtual audio cable. Windows: VB-CABLE from https://vb-audio.com/Cable/ (I run its installer as
     administrator and restart). Mac: brew install --cask blackhole-2ch (it asks for my password).
     Only from these places.
   - uv (https://docs.astral.sh/uv/getting-started/installation/).
3. In tts/, make the voice environment: uv venv -p 3.11, then uv pip install -r requirements-kokoro.txt
   (on a PC with an NVIDIA card add --extra-index-url https://download.pytorch.org/whl/cu128).
4. Run uv run python server.py --list-devices in tts/ and show me the output devices.

End with a checklist: Unreal 5.8, OBS, virtual cable, voice environment, each with a tick or a cross.

Vous devriez voir : quatre coches : Unreal 5.8, OBS, le câble et la voix.

Open the kit's Unreal project and connect to Unreal's own MCP server.

1. Open FaceNinjaCreatorKit/FaceNinjaCreatorKit.uproject with Unreal Engine 5.8 (Mac: open -a with
   "/Users/Shared/Epic Games/UE_5.8/Engine/Binaries/Mac/UnrealEditor.app" and the project path; Windows:
   "C:\Program Files\Epic Games\UE_5.8\Engine\Binaries\Win64\UnrealEditor.exe" with the project path).
   Tell me the first start takes several minutes while it compiles shaders.
2. The project starts Unreal's MCP server at http://127.0.0.1:8000/mcp. This folder already has the MCP
   settings for you (.mcp.json, .cursor, .vscode, .gemini, .codex). If you cannot see an "unreal-mcp" server,
   tell me how to reload your MCP settings, or add that URL yourself.
3. Wait until the server answers, then list the toolsets. You should find FaceNinjaKitTools. Describe it and
   tell me in one sentence per tool what you can now do for me.

Vous devriez voir : votre agent liste les outils FaceNinja.

Using FaceNinjaKitTools in Unreal, build my streaming stage.

1. Call build_stage. It makes a small room with a sofa, a bookshelf, a plant, lights and three cameras.
2. Take a still from each camera with stage_screenshot (closeup, bust, wide) and show them to me.
3. Ask me which camera I want for streaming and call set_camera with it.

Vous devriez voir : trois images fixes de la pièce.

Using FaceNinjaKitTools in Unreal, turn my FaceNinja head into a MetaHuman and put it on my stage.
My head ZIP is: <drag the ZIP file here>

1. Call import_head_zip with that path. Tell me which hair, brows and lashes it picked.
2. Tell me that the next step sends my head's shape (not my photos) to Epic's free MetaHuman service, and that
   a browser window may ask me to sign in to my Epic account. Wait for my OK.
3. Call preview_avatar so I can see my head in the room, and show me a stage_screenshot (bust camera).
4. Call rig_head, then download_head_textures. Tell me: "A browser opens Epic's sign-in page. Unreal stops
   answering until you sign in and approve it there; that is normal." Wait until I say I approved it, then
   check head_status every 30 seconds until rigged and high_res_textures are both true.
5. Once rigged: call build_metahuman, then place_avatar.
6. Take a stage_screenshot with the bust camera and show it to me.

Vous devriez voir : votre MetaHuman dans la pièce.

Using FaceNinjaKitTools in Unreal, connect my iPhone's Live Link Face app to my avatar.

1. Call my_ip_address. Tell me where to type it in the app (Settings > Live Link > Add Target, port 14785).
2. Ask me for the phone's IP address (Live Link Face shows it in its settings), then call connect_iphone.
3. Call connect_face with FN_iPhone, then play_in_window.
4. Ask me to smile, blink and turn my head, and whether the avatar follows. Ask before changing any firewall
   setting.

Vous devriez voir : l’avatar copie votre visage.

Start the kit's voice server and connect it to my avatar.

1. Ask me which voice I want:
   a. a free preset voice that runs on my computer (Kokoro, default),
   b. my own voice (Chatterbox): ask me to record 10-20 seconds of myself reading any text in a quiet room as
      tts/my_voice.wav, then set it up with requirements-chatterbox.txt,
   c. a studio voice from ElevenLabs: it needs my own ElevenLabs account, costs money per line and sends my text to ElevenLabs.
      Ask me to put my own ElevenLabs API key in the ELEVENLABS_API_KEY environment variable myself; never write it to a file.
2. In tts/ start the server in the background with my cable as --device ("BlackHole" on a Mac, "CABLE Input" on
   Windows):
   a. uv run python server.py --engine kokoro --preset af_heart --device "<cable>"
   b. uv run python server.py --engine chatterbox-turbo --voice my_voice.wav --device "<cable>"
   c. uv run python server.py --engine elevenlabs --voice Rachel --device "<cable>"
3. In Unreal, using FaceNinjaKitTools: voice_status, then connect_voice (it listens on the cable), then
   connect_face with FN_Voice, then play_in_window, then open_talk_panel.
4. Call say with "Hi everyone, welcome back to the stream!" and ask me if the lips moved with the voice.

Vous devriez voir : il dit votre phrase en bougeant les lèvres.

Help me show my avatar in OBS Studio.

1. In Unreal, using FaceNinjaKitTools: play_in_window (1920x1080). If I want my own background instead of the
   room, call set_green_screen with true first.
2. In OBS add a "Window Capture" (Windows) or "macOS Screen Capture" set to that Unreal window. Tell me the
   clicks; you cannot click in OBS for me.
3. Only with the green screen: add a "Chroma Key" filter (key colour green) and show me where "Similarity" is.
4. Add an "Audio Input Capture" of "CABLE Output" (Windows) or "BlackHole 2ch" (Mac), and set its
   "Audio Monitoring" to "Monitor and Output" so I hear the voice too.

Vous devriez voir : votre avatar dans OBS, avec sa voix.