Installation locale PPT Master (modifié) — Guide complet
Prérequis
- Windows 10/11 avec une RTX 5070 Ti
- Python 3.10+ installé (https://www.python.org/downloads/) → cocher "Add to PATH"
- Git installé (https://git-scm.com/downloads)
- ~15 Go d'espace disque (dont ~6 Go pour un checkpoint SDXL)
Étape 1 : Cloner PPT Master
Ouvre un terminal (PowerShell ou CMD) :
cd C:\Users\TON_USER\Documents
git clone https://github.com/hugohe3/ppt-master.git
cd ppt-master
pip install -r requirements.txt
Étape 2 : Installer ComfyUI
cd C:\Users\TON_USER\Documents
git clone https://github.com/comfyanonymous/ComfyUI.git
cd ComfyUI
pip install -r requirements.txt
pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu128
Vérifie que ta RTX est détectée :
python -c "import torch; print(torch.cuda.is_available()); print(torch.cuda.get_device_name(0))"
Résultat attendu :
True
NVIDIA GeForce RTX 5070 Ti
Étape 3 : Télécharger un modèle image
Télécharge un checkpoint Stable Diffusion dans ComfyUI/models/checkpoints/ :
Option A (recommandée, qualité top) :
- Juggernaut XL : https://civitai.com/models/133005/juggernaut-xl
- Télécharge le fichier
.safetensorset place-le dansComfyUI/models/checkpoints/
Option B (officiel, plus léger) :
- SDXL Base : https://huggingface.co/stabilityai/stable-diffusion-xl-base-1.0/resolve/main/sd_xl_base_1.0.safetensors
Étape 4 : Appliquer les modifications PPT Master
Copie les fichiers du dossier modifications/ dans ton clone PPT Master :
modifications/backend_comfyui.py → ppt-master/skills/ppt-master/scripts/image_backends/backend_comfyui.py
modifications/openrouter_llm.py → ppt-master/skills/ppt-master/scripts/openrouter_llm.py
modifications/.env → ppt-master/.env
Puis édite ppt-master/.env et ajuste les chemins :
COMFYUI_DIR=C:\Users\TON_USER\Documents\ComfyUI
COMFYUI_MODEL=juggernautXL_v9.safetensors
COMFYUI_FALLBACK=error
Étape 5 : Modifier image_gen.py
Ouvre ppt-master/skills/ppt-master/scripts/image_gen.py et ajoute le backend ComfyUI.
Dans la section IMAGE_ENV_PREFIXES, ajoute "COMFYUI_", à la fin de la liste.
Dans le dictionnaire BACKEND_REGISTRY, ajoute avant "openrouter" :
"comfyui": {
"module": "backend_comfyui",
"tier": "core",
"label": "ComfyUI Local",
"default_model": "local-comfyui",
"key_hint": "COMFYUI_SERVER / COMFYUI_WORKFLOW / COMFYUI_MODEL",
"aliases": ["comfy", "local"],
},
Étape 6 : Tester
Test ComfyUI seul (lance-le d'abord) :
cd C:\Users\TON_USER\Documents\ComfyUI
python main.py --listen 127.0.0.1 --port 8188
Tu dois voir dans la console : To see the GUI go to: http://127.0.0.1:8188
Test génération d'image :
Dans un autre terminal :
cd C:\Users\TON_USER\Documents\ppt-master
python skills/ppt-master/scripts/image_gen.py "A cinematic dark fantasy throne room" --backend comfyui --aspect_ratio 16:9 --image_size 1K -o test_images
Test LLM OpenRouter :
cd C:\Users\TON_USER\Documents\ppt-master
python skills/ppt-master/scripts/openrouter_llm.py "Dis bonjour"
Étape 7 : Utiliser PPT Master normalement
PPT Master est conçu pour fonctionner dans un IDE agent (Claude Code, Cursor, etc.).
L'agent lit skills/ppt-master/SKILL.md et suit le workflow.
Mais avec les modifications, tu peux aussi l'utiliser "à la main" en suivant le pipeline :
python scripts/project_manager.py init mon_projet --format ppt169- Écrire
design_spec.mdetspec_lock.md(ou utiliser OpenRouter pour les générer) - Générer les images :
python scripts/image_gen.py "prompt" --backend comfyui -o mon_projet/images - Créer les SVG dans
svg_output/ python scripts/total_md_split.py mon_projetpython scripts/finalize_svg.py mon_projetpython scripts/svg_to_pptx.py mon_projet
Résumé des fichiers modifiés/ajoutés
| Fichier | Action |
|---|---|
.env |
NOUVEAU — config OpenRouter + ComfyUI |
scripts/image_backends/backend_comfyui.py |
NOUVEAU — backend ComfyUI local |
scripts/openrouter_llm.py |
NOUVEAU — client LLM OpenRouter |
scripts/image_gen.py |
MODIFIÉ — ajout du backend comfyui au registre |
Dépannage
| Problème | Solution |
|---|---|
torch.cuda.is_available() = False |
Réinstalle PyTorch avec CUDA : pip install torch --index-url https://download.pytorch.org/whl/cu128 |
| ComfyUI ne démarre pas | Vérifie que le port 8188 est libre, que Python trouve torch |
| Image placeholder au lieu de vraie image | Vérifie COMFYUI_MODEL dans .env et que le fichier existe dans ComfyUI/models/checkpoints/ |
| OpenRouter erreur 401 | La clé API est peut-être expirée, vérifie sur openrouter.ai |
image_gen.py ne trouve pas comfyui |
Vérifie que tu as bien ajouté l'entrée dans BACKEND_REGISTRY de image_gen.py |