Hey Pinokio community! It has been an incredible 2 months since our initial v1.1.0 release. Thanks to your feedback and testing, **Companion for ComfyUI** has evolved from a utility app into a full-scale, enterprise-ready **v2.0.1 Desktop Studio**!
Whether you are running ComfyUI on an RTX 4090 or a modest 8GB/16GB VRAM card, v2.0.1 brings groundbreaking speed improvements, intelligent hardware management, and deep workflow integration.
---
What's New Since the Last Update
1. Civitai Model Hub & 1-Click Canvas Bridge
* **In-App Model Discovery:** Browse trending checkpoints, LoRAs, and embeddings from Civitai directly inside the Companion app.
* **Dual Action Modes:** Send runnable LiteGraph workflow templates straight to your active ComfyUI canvas with 1 click, or trigger instant model downloads with automatic folder placement.
* **Smart Dependency Resolver:** Missing base models or CLIP/VAE files are automatically mapped and cued for download.
---
2. Video Architect 2.0 (Local AI Scriptwriting & Lip Sync)
* **Upgraded Local LLMs (Gemma-3-4B & Qwen-2.5-7B):** Run smart storyboard deconstruction and visual prompt engineering 100% locally on your machine with zero cloud dependence.
* **⚡ 16 GB VRAM Optimization Pipeline:**
- Automated `VAEDecode` tiled scaling (512px tiles) prevents out-of-memory crashes during high-res rendering.
- Automatic `CLIPLoader` offloading to CPU frees **6–8 GB of precious VRAM**.
- Quantized `fp8_e4m3fn` weights enforced on UNET loaders for buttery-smooth video generation.
* **Audio-Driven Lip Sync & Head Animation:** Seamless pipeline integration with SadTalker, LivePortrait, and Wav2Lip.
* **Two-Phase Rendering & Last-Frame Chaining:** Guarantees temporal consistency and smooth scene transitions with Ken Burns cinematic motion, Piper TTS / ComfyUI voiceovers, and automatic subtitle generation.
---
3. Non-Destructive Mode & Auto-Discovery (Addressing Community Feedback!)
* **Zero-Risk to Existing Setups:** Answering questions from our Pinokio thread: Companion **never mutates or overwrites** your existing Python packages, CUDA environments, or custom attention kernels (`flash-attn`, `sage2`, `triton`).
* **First-Run Setup Wizard:** Automatically scans existing ComfyUI directories (standalone, WSL, or existing Pinokio apps) and imports `extra_model_paths.yaml` paths without moving your files.
* **Unified Portable / AppData Support:** Full support for running in portable directory mode alongside Pinokio.
---
4. High-Speed Segmented Multi-Stream Downloader (v2.0.1)
* **Multi-Segment HTTP Streaming:** Download massive 10GB–25GB model files in parallel chunks using Java 21 Virtual Threads.
* **Zero-Copy NIO.2 FileChannel Stitching:** Slices are merged directly on disk with zero memory overhead.
* **Smart Boundary Cutoffs & Auto-Fallback:** Solves HTTP keep-alive connection stalls on Cloudflare/Backblaze CDN endpoints with strict HTTP 206 validation and seamless automatic fallback to single-stream resume.
---
5. Prompt Lab Superpowers: XY Matrix, PNG Remix & History Recall
* **XY Prompt Matrix Grid:** Systematically benchmark prompts, CFG scales, step counts, and samplers against each other in an automated visual grid.
* **PNG Metadata Remix:** Drag & drop any generated image to instantly recall all original generation settings, prompt syntax, seed, and sampler parameters.
* **VRAM Traffic-Light Advisor:** A live heuristic indicator alerts you before generation whether a workflow is safe (Green), tight (Yellow), or likely to trigger an out-of-memory crash (Red).
* **Prompt History Recall:** Full indexed history of your generated prompts with instant one-click restoration.
---
6. Storage Cleaner, NTFS Hardlinks & Snapshot Rollback
* **Duplicate Merging:** Identifies duplicate model weights using AutoV1 and SHA-256 and merges them via NTFS hardlinks, saving gigabytes of SSD storage while maintaining existing folder paths.
* **Cold Storage Archive:** Effortlessly offload rarely used models to external drives and restore them with 1 click when a workflow calls for them.
* **Environment Snapshots:** Create snapshot points of your configuration for instant rollback.
---
How to Install & Update on Pinokio
If you already have Companion installed in Pinokio:
1. Open **Pinokio**.
2. Click **Companion for ComfyUI**.
3. Click **Update** (this runs `git pull` and re-packages the latest v2.0.1 JAR automatically).
4. Click **Start**!
**New to Companion?**
* Install directly via the Pinokio app page:
**[https://pinokio.co/apps/github-com-thomaskippster-comfyuicompanion](https://pinokio.co/apps/github-com-thomaskippster-comfyuicompanion)**
* Or paste the repo URL into Pinokio's search bar:
`https://github.com/thomaskippster/comfyuicompanion\`
---
*Open-source, local-first, and built for creators. Enjoy v2.0.1!* 🚀
Hey Pinokio community! It has been an incredible 2 months since our initial v1.1.0 release. Thanks to your feedback and testing, **Companion for ComfyUI** has evolved from a utility app into a full-scale, enterprise-ready **v2.0.1 Desktop Studio**!
Whether you are running ComfyUI on an RTX 4090 or a modest 8GB/16GB VRAM card, v2.0.1 brings groundbreaking speed improvements, intelligent hardware management, and deep workflow integration.
---
What's New Since the Last Update
1. Civitai Model Hub & 1-Click Canvas Bridge
* **In-App Model Discovery:** Browse trending checkpoints, LoRAs, and embeddings from Civitai directly inside the Companion app.
* **Dual Action Modes:** Send runnable LiteGraph workflow templates straight to your active ComfyUI canvas with 1 click, or trigger instant model downloads with automatic folder placement.
* **Smart Dependency Resolver:** Missing base models or CLIP/VAE files are automatically mapped and cued for download.
---
2. Video Architect 2.0 (Local AI Scriptwriting & Lip Sync)
* **Upgraded Local LLMs (Gemma-3-4B & Qwen-2.5-7B):** Run smart storyboard deconstruction and visual prompt engineering 100% locally on your machine with zero cloud dependence.
* **⚡ 16 GB VRAM Optimization Pipeline:**
- Automated `VAEDecode` tiled scaling (512px tiles) prevents out-of-memory crashes during high-res rendering.
- Automatic `CLIPLoader` offloading to CPU frees **6–8 GB of precious VRAM**.
- Quantized `fp8_e4m3fn` weights enforced on UNET loaders for buttery-smooth video generation.
* **Audio-Driven Lip Sync & Head Animation:** Seamless pipeline integration with SadTalker, LivePortrait, and Wav2Lip.
* **Two-Phase Rendering & Last-Frame Chaining:** Guarantees temporal consistency and smooth scene transitions with Ken Burns cinematic motion, Piper TTS / ComfyUI voiceovers, and automatic subtitle generation.
---
3. Non-Destructive Mode & Auto-Discovery (Addressing Community Feedback!)
* **Zero-Risk to Existing Setups:** Answering questions from our Pinokio thread: Companion **never mutates or overwrites** your existing Python packages, CUDA environments, or custom attention kernels (`flash-attn`, `sage2`, `triton`).
* **First-Run Setup Wizard:** Automatically scans existing ComfyUI directories (standalone, WSL, or existing Pinokio apps) and imports `extra_model_paths.yaml` paths without moving your files.
* **Unified Portable / AppData Support:** Full support for running in portable directory mode alongside Pinokio.
---
4. High-Speed Segmented Multi-Stream Downloader (v2.0.1)
* **Multi-Segment HTTP Streaming:** Download massive 10GB–25GB model files in parallel chunks using Java 21 Virtual Threads.
* **Zero-Copy NIO.2 FileChannel Stitching:** Slices are merged directly on disk with zero memory overhead.
* **Smart Boundary Cutoffs & Auto-Fallback:** Solves HTTP keep-alive connection stalls on Cloudflare/Backblaze CDN endpoints with strict HTTP 206 validation and seamless automatic fallback to single-stream resume.
---
5. Prompt Lab Superpowers: XY Matrix, PNG Remix & History Recall
* **XY Prompt Matrix Grid:** Systematically benchmark prompts, CFG scales, step counts, and samplers against each other in an automated visual grid.
* **PNG Metadata Remix:** Drag & drop any generated image to instantly recall all original generation settings, prompt syntax, seed, and sampler parameters.
* **VRAM Traffic-Light Advisor:** A live heuristic indicator alerts you before generation whether a workflow is safe (Green), tight (Yellow), or likely to trigger an out-of-memory crash (Red).
* **Prompt History Recall:** Full indexed history of your generated prompts with instant one-click restoration.
---
6. Storage Cleaner, NTFS Hardlinks & Snapshot Rollback
* **Duplicate Merging:** Identifies duplicate model weights using AutoV1 and SHA-256 and merges them via NTFS hardlinks, saving gigabytes of SSD storage while maintaining existing folder paths.
* **Cold Storage Archive:** Effortlessly offload rarely used models to external drives and restore them with 1 click when a workflow calls for them.
* **Environment Snapshots:** Create snapshot points of your configuration for instant rollback.
---
How to Install & Update on Pinokio
If you already have Companion installed in Pinokio:
1. Open **Pinokio**.
2. Click **Companion for ComfyUI**.
3. Click **Update** (this runs `git pull` and re-packages the latest v2.0.1 JAR automatically).
4. Click **Start**!
**New to Companion?**
* Install directly via the Pinokio app page:
**[https://pinokio.co/apps/github-com-thomaskippster-comfyuicompanion](https://pinokio.co/apps/github-com-thomaskippster-comfyuicompanion)**
* Or paste the repo URL into Pinokio's search bar:
`https://github.com/thomaskippster/comfyuicompanion\`
---
*Open-source, local-first, and built for creators. Enjoy v2.0.1!* 🚀