English · Français

Windows RTX 5090 Cinema Workstation - Full Install Log

Date: 8 mars 2026, 00h27 - 03h35 (Session 1: 27 min + Session 2: ~2h30 ongoing) Machine: Windows 11, NVIDIA GeForce RTX 5090 32GB VRAM, Driver 591.74 Method: Claude Code (Opus 4.6) via Pinokio - fully automated, zero manual intervention Operator: Ismael Joffroy Chandoutis (@ismael-joffroy-chandoutis)

---

The Numbers

Category	Count	Time
Pinokio apps/repos cloned	146	~20 min
Windows apps installed (winget)	72	~20 min (parallel)
Python packages installed	164	~5 min
Node.js global packages	6	~1 min
Ollama LLM models	6	~30 min (parallel)
ComfyUI custom nodes	95	~10 min
ComfyUI models (checkpoints, upscalers, etc.)	22+	~15 min
LoRA training tools	5	~5 min
Silent installers executed	2	~3 min
TOTAL installations	~520+	Session 1: 27 min, Session 2: ongoing

---

Method

Everything was installed by a single Claude Code (Opus 4.6) session running inside Pinokio on Windows. The process:

1. Connected to GitHub via gh auth login --web (browser auth)
2. Scanned all user repos (47) and starred repos (200+)
3. Analyzed user profile to understand needs (award-winning documentary filmmaker)
4. Launched parallel installations using winget, git clone, pip, npm, ollama
5. Used Pinokio's pterm CLI to trigger install.js scripts for apps requiring venv/PyTorch setup
6. No manual intervention at any point after initial GitHub auth

Key techniques:

• Massive parallelization: 5-10 background tasks running simultaneously
• winget for Windows desktop apps (silent install, no GUI interaction)
• git clone directly into C:/pinokio/api/ for Pinokio app discovery
• pterm start for Pinokio-native installations (venv, PyTorch, CUDA)
• ollama pull for LLM model downloads in background

---

Complete Software Inventory

Desktop Applications (via winget)

App	Version	Category
Blender	5.0.1	3D / VFX
DaVinci Resolve	Latest	Video Editing / Color
Kdenlive	Latest	Video Editing
Shotcut	Latest	Video Editing
OBS Studio	32.0.4	Screen Recording
Audacity	3.7.7	Audio Editing
Krita	5.2.16	Digital Painting / AI
GIMP	Latest	Image Editing
Inkscape	1.4.3	Vector Graphics
Godot Engine	4.6.1	Game Engine / Interactive
VLC	3.0.23	Media Player
mpv.net	7.1.2	Media Player (pro)
HandBrake	1.10.2	Video Transcoding
FFmpeg	8.0.1	Video/Audio Processing
ImageMagick	7.1.2	Image Processing
Shutter Encoder	18.9	Professional Encoder
DCP-o-matic	Latest	DCP Creation (cinema distribution)
Subtitle Edit	4.0.15	Subtitle Editor
MediaInfo	26.01	Media Analysis
Kyno	1.9.0.2	Media Browser / Manager
XnView MP	1.10.3	Image Viewer / Manager
ExifTool	Latest	Metadata Tool
LM Studio	Latest	Local LLM Interface
Ollama	0.17.6	Local LLM Server
Epic Games Launcher	1.3.173	For Unreal Engine 5
VS Code	1.110	Code Editor
Notepad++	8.9.2	Text Editor
Obsidian	1.12.4	Markdown / Knowledge Base
Everything	1.4.1	File Search
7-Zip	26.00	Archive Manager
Tailscale	1.94.2	Mesh VPN
Telegram	6.6.2	Messaging
qBittorrent	5.1.4	Torrent Client
Termius	9.37.2	SSH Client
PuTTY	0.83	SSH Client
WinSCP	6.5.5	SFTP Client
yt-dlp	2026.03.03	Video Downloader
Google Cloud SDK	Latest	Cloud CLI
AnyDesk	9.6.9	Remote Desktop
Docker Desktop	4.61.0	Containers

Pre-existing: Adobe Creative Cloud, Chrome, Edge, Brave, Discord, Slack, WhatsApp, Instagram, Canva, CapCut, Claude Desktop, ChatGPT, Notion, Perplexity, Bitwarden, 4K Video Downloader+, Apple TV/Devices/iCloud, Microsoft Teams, Parsec, Deno

Pinokio AI/Creative Apps (146 repos)

Session 2 additions: 69 official Pinokio apps + ComfyUI ecosystem expansion

AI Image/Video Generation

App	Source	Description
ComfyUI	comfy.git	Main AI generation hub
Forge Neo	6Morpheus6	SD WebUI Forge - Flux, Z-Image, Wan, Lumina, Kontext, nunchaku
Forge Classic	Haoming02	Updated SD WebUI Forge
Z-Fusion	ai-anchorite	Z-Image, Flux2 Klein, SeedVR2 with Gradio
Z-Image-Pinokio	PierrunoYT	Z-Image Turbo - sub-second inference
wan2gp	6Morpheus6	Wan 2.1/2.2, Qwen, HunyuanVideo, LTX, Flux, OVI
Frame-Pack	pinokiofactory	Progressive video generation
fp-studio	FP-Studio	FramePack enhanced - LoRA, upscaling, interpolation
LTX-Desktop	Lightricks	LTX Video desktop app
Open-Higgsfield-AI	Anil-matcha	Open-source cinema studio AI
comfyui-photoshop	NimaNzrii	ComfyUI inside Photoshop
FlashVSR	ai-anchorite	Video Super Resolution upscaler
InvokeAI	eudard	Alternative generation pipeline
SwarmUI	SUP3RMASS1VE	SwarmUI interface
ai-toolkit	ai-anchorite	Training/finetuning diffusion models
ComfyDock	comfygit-ai	Docker-based ComfyUI environments
FaceFusion	facefusion	Face manipulation platform
roop.pinokio	AmusedDiffuser	Face swapping
roop-unleashed	SUP3RMASS1VE	Advanced face swapping
Deepseek Janus Pro 7B	SUP3RMASS1VE	Multimodal AI
deeperhermes	cocktailpeanut	Self-deciding LLM
clarity-refiners-ui	pinokiofactory	Image upscaler/enhancer
krita-ai-diffusion	Acly	AI generation inside Krita

Audio / TTS / Music

App	Source	Description
AllTalk-TTS	6Morpheus6	F5, XTTS, Vite, Piper, Parler, RVC
Ultimate TTS Studio	SUP3RMASS1VE	Kokoro, KittenTTS, Chatterbox, Fish, F5, IndexTTS
TTS-Story	Xerophayze	Multi-voice audiobook studio
Alexandria Audiobook	Finrandojin	Audiobook generator with Qwen3-TTS
Qwen3-TTS-Pinokio	SUP3RMASS1VE	Qwen3 TTS
Orpheus-TTS	pinokiofactory	Orpheus TTS FastAPI
StyleTTS2 Studio	pinokiofactory	Custom voice building
KittenTTS	soldier444xd	Ultra-lightweight TTS (15M params)
Chattered	6Morpheus6	Chatterbox voice cloning
SongGeneration-Studio	BazedFrog	Song generation from text/audio
Audiocraft Plus	cocktailpeanut	MusicGen + AudioGen
StemXtract	TheAwaken1	Audio stem separation (vocals/drums/bass)
MMAudio	pinokiofactory	Multimodal audio generation
Parakeet-TDT	SUP3RMASS1VE	Speech recognition
Whisper WebUI	pinokiofactory	Transcription (Whisper)

3D / Upscaling

App	Source	Description
TRELLIS	pinokiofactory	3D Gaussian Splatting generation

AI Workflows / Automation

App	Source	Description
Flowise	FlowiseAI	Visual AI workflow builder
bolt.new	gotoolkits	AI-powered web dev in browser
ai-file-sorter	hyperfield	Content-aware file organization

Personal Projects (22 repos)

Repo	Description
goldberg-database	The Goldberg Variations - documentary data
goldberg-visu	Unlisted visualizations
buttercut	Edit video with Claude Code
cc-wf-studio	CC Workflow Studio
cinema-ai-toolkit	Documentary filmmaker's AI toolkit
comfyui-cinema-pipeline	70+ ComfyUI workflows for cinema
comfyui-blender-temporal	Blender EXR to ControlNet nodes
storyexplorer	Nonlinear narrative exploration
storycurve	Story curve visualization
agent-viewer	Kanban board for Claude Code agents
ai-cinema-method	AI as artistic method - research
filmmaker-tools	Claude Code plugin for filmmakers
filmmaker-ai-brain	Cinema memory architecture
vaisseau-amiral	Personal intelligence system
claude-config	Claude configuration
claude-code-filmmaker	How a filmmaker uses Claude Code
claude-skills	Custom Claude Code skills
claude-system-prompts	System prompts and templates
claude-agents	Autonomous agents and workflows
claude-mcp-servers	MCP server integrations
claude-commands	Custom slash commands
IJC_Claude_Code	Claude Code workspace
deep-research	Research and documentation
articles	Articles and writing
joshua-post-timeline	Post-timeline visualization
web-apps	Web applications
python-tools	Python scripts and automation
blender-ai-workflows	Blender AI filmmaking workflows
cinema-tools	Film analysis tools
open-source-cinema	RAW video filmmaking
n8n-automation-ijc	n8n automation setup
postiz-setup	Postiz social media setup
vibe-ribbon	3D avatar web app
mcp-server-tmdb	MCP server for TMDB

LLM Models (Ollama)

Model	Size	Use Case
Qwen3.5-32B	~20GB	Best creative writing FR+EN, SOTA mars 2026
Qwen3.5-14B	~9GB	Fast creative, excellent multilingue
DeepSeek R1 32B	~20GB	Chain-of-thought reasoning
Mistral-Nemo 12B	~8GB	Excellent français natif
Dolphin3-abliterated	~5GB	Uncensored creative (scénario gore/sexuel)
Mistral-Nemo-abliterated	~8GB	Uncensored français

ComfyUI Custom Nodes (95 nodes)

Core & Essential

ComfyUI-Manager, Impact Pack, IPAdapter Plus, WAS Node Suite, Essentials, Custom Scripts, KJNodes, Crystools, Easy-Use, rgthree, Comfyroll, workspace-manager

Image Generation & Control

SUPIR (upscaler), CCSR, x-flux, Flux2Wrapper, Z-Image, Flux2Fun ControlNet, ControlAltAI (Flux Union), FluxTrainer, Fluxtapoz (Kontext), PuLID (face ID), InstantID, PixArt, layerdiffuse, SeeCoder, sdxl_prompt_styler, Shakker Nodes

Video Generation

WanVideoWrapper, Wan2.2Wrapper, HunyuanVideoWrapper, CogVideoXWrapper, LTXVideo, LTXTricks, FramePackWrapper, DynamiCrafterWrapper, AnimateDiff Evolved, Steerable Motion, TemporalKit, Frame Interpolation, Optical Flow, VideoHelperSuite, LatentSyncWrapper, PainterLongVideo, RAVE, StableAudioX

ControlNet & Preprocessing

Advanced ControlNet, DepthAnythingV2, Marigold, controlnet_aux, SAM2, segment_anything, Florence2, BiRefNet

Face & Portrait

reactor (face swap), FaceAnalysis, facerestore, LivePortraitKJ, AdvancedLivePortrait, IC-Light

Upscaling

UltimateSDUpscale, TiledDiffusion, TiledKSampler, SeedVR2 VideoUpscaler, image-resize

3D

Hunyuan3D-2.1, Hunyuan3DWrapper, StableCascade

Animation & Effects

Deforum, Deforum-X-Flux, Disco Diffusion, ArtGallery, Color Transfer, Latent Modifiers, differential-diffusion, Noise, FizzNodes, LayerForge

Utilities

anynode, cg-use-everywhere, cg-image-picker, Image Saver, prompt-control, prompt-reader-node, ProfilerX, yaResolutionSelector, masquerade, sd-dynamic-thresholding, Inpaint-CropAndStitch, GGUF, CivitAI, Qwen-VL-API, nunchaku-nodes, efficiency-nodes, stable-audio-tools

LoRA Training Tools (5)

Tool	Strength
kohya_ss	Reference avancée, meilleure généralisation
SimpleTuner	Précision low VRAM (Optimum-Quanto)
OneTrainer	Masked training, text-encoder
FluxGym	GUI simple (Kohya + AI-Toolkit)
ai-toolkit (Ostris)	Le plus simple pour Flux

ComfyUI Models

Model	Type	Size
Flux 1 Dev BnB NF4	Checkpoint	12GB
Flux 1 Schnell FP8	Checkpoint	11GB
Flux 2 Dev FP8 Mixed	Diffusion	downloading
Flux 2 Turbo LoRA	LoRA	downloading
Flux 2 Klein 4B	Diffusion	downloading
Flux 2 VAE + Text Encoders	VAE/TE	done
Z-Image BF16 + VAE + TE	Full stack	downloading
LTX 2.3 Distilled	Video	downloading
LTX 2.3 Spatial/Temporal Upscaler	Upscaler	downloading
HunyuanVideo 720 FP8	Video	13GB
FramePack I2V HY FP8	Video	done
Wan 2.2 Fun Control 14B FP8	Video	done
Wan 2.2 I2V 14B FP8	Video	done
Flux 1 Kontext FP8	Editing	done
Flux 1 Fill Dev	Inpainting	done
Qwen Image 7B FP8	Image Gen	done
ACE-Step v1 3.5B	Music	done
SD 1.5 Pruned	Checkpoint	downloading
SDXL Base 1.0	Checkpoint	downloading
SDXL Refiner 1.0	Checkpoint	5.7GB
4x-UltraSharp	Upscaler	64MB
RealESRGAN x4plus	Upscaler	64MB
RealESRGAN x4plus anime	Upscaler	18MB
ControlNet ProMax	ControlNet	2.4GB

Python Packages (164 total, key ones)

• AI/ML: anthropic, openai, google-generativeai, replicate, huggingface-hub, sentence-transformers, langchain
• Vector DB: pinecone-client, chromadb
• Media: gallery-dl, spotdl, whisper-ctranslate2, pydavinci
• Dev: uv, pip, setuptools

Node.js Global Packages

• @anthropic-ai/sdk, @google/generative-ai, marked-terminal

---

What Makes This Insane

1. 520+ total installations across 8+ different package managers/methods
2. Zero manual clicks - everything automated via Claude Code CLI
3. Massive parallelization - up to 15 concurrent background installs
4. Intelligent profiling - Claude analyzed GitHub repos to determine exact needs
5. Cross-ecosystem - Windows apps, Python venvs, Node packages, Ollama models, Pinokio scripts, HuggingFace models, ComfyUI nodes
6. Production-grade - professional cinema/AI tools, not toy installs
7. RTX 5090 optimized - PyTorch CUDA 13.0, xformers, proper GPU acceleration
8. SOTA March 2026 - Flux 2, LTX 2.3, Z-Image, Qwen 3.5, Wan 2.2, latest everything
9. Shared model architecture - Pinokio Drive symlinks = zero duplication across interfaces
10. Training-ready - 5 LoRA trainers, FluxTrainer node, full training pipeline

The Pipeline

Pinokio (app launcher)
  -> Claude Code (Opus 4.6, orchestrator)
    -> GitHub CLI (auth + repo discovery)
    -> winget (Windows apps, parallel)
    -> git clone (Pinokio apps, batch)
    -> pip/uv (Python packages)
    -> npm (Node packages)
    -> ollama (LLM models, full path)
    -> huggingface_hub (model downloads)
    -> pterm (Pinokio install scripts)
    -> curl (direct downloads)
    -> cmd.exe /S (silent installers)
    -> cp -r (Pinokio official examples)

This is what happens when you give an AI agent full system access and tell it to go. The entire Windows machine went from a bare Pinokio + ComfyUI setup to a complete professional filmmaker's AI workstation - the kind of setup that would take a human team days to configure.

---

By Ismaël Joffroy Chandoutis.