DruidCat's ComfyUI Workflows, Scripts and Templates for Runpod

Aiya! I've just released the most comprehensive AI influencer creation course ever made. This is literally months of development packed into one ultimate open-source collection with 40+ professional workflows for creating realistic AI influencers.

Full tutorial here:

https://youtu.be/igEvZtVX8j0

ComfyUI installation scripts working both on local and Runpod (cloud rental GPUs) packed with improvements and brand new assets — including installation scripts and enhanced workflows (custom-modified)!

The installation scripts work exactly the same way as shown in the https://youtu.be/n_mAPAjc4Fs and the workflows are thoroughly documented inside — explaining step by step what each node does, guiding the user "by the hand." For those already familiar with ComfyUI, they should be easy to use.

What's inside?

To install a specific workflow, simply navigate into the druidcat_scripts, then to the folder of one you want to install, then open a terminal (Runpod, Ubuntu) or PowerShell window (you can use cd "path to your folder" to navigate) :

• If you're using the terminal on RunPod, enter:bash install.sh
• If you're on Windows using PowerShell, run:Set-ExecutionPolicy Bypass -Scope Process -Force; .\install.ps1

✅ Fixes & Improvements

• Squashed multiple bugs that were affecting earlier versions.
• Improved installers: the template now includes two comfyhub versions — comfyhub and comfyhub 5090. I attached pdf files.
• The comfyhub 5090 version works only with NVIDIA RTX 5090 GPUs, so if you're running in the cloud with an RTX 5090, you can now generate art and videos at even faster speeds than before.

🆕 Tons of New Workflows & ModelsYou’ll find all the latest additions that have emerged since last month’s rollout:

• Workflows for Flux, Sonic, Upscalers, HiDream and LTX Studio
• Flux Kontext Dev & Kontext Dev Change Clothes — powerful FLUX-context tools for consistent character style
• WAN 2.1 FusionX — optimized for fast, high-fidelity unlimited length video generation (crazy good for Tik Tok), and also includes additional Text‑to‑Image workflows (better then FLUX).
• WAN 2.2 — a major upgrade, supporting Text‑to‑Video, Image‑to‑Video (better then WAN 2.1)., and the innovative First‑Frame/Last‑Frame (FLF2V) cinematic pipeline with speed, frame interpolation and upscaling improvments
• WAN Multitalk — create talking avatars and animated portraits

⚙️ Local Installation Notes & Scripts

I have improved the pre-install dependency script for Windows. Remember to install Git and Python with the “Add to PATH” option enabled.

Go to druidcat_scripts unzipped folder. Run the setup in PowerShell with execution policy disabled using this command:

Set-ExecutionPolicy Bypass -Scope Process -Force; .\install_dependencies_Windows.ps1

🧠 Fine‑Tuning / LoRA Training Setup (new template)Many users reported the old LoRA training template would hang. With WAN support now added, I’m switching to a new pipeline based on Ostris’s template AI tools on Runpod https://console.runpod.io/explore/0fqzfjy6f3 

This comes with AI tools tips and step-by-step guides for training WAN‑based models for text to image.

• You can google tutorials from the same developer for training best practices, and I’ll include them in my upcoming tutorial. https://www.youtube.com/@ostrisai

Unfortunately, Ostris's AI tools don't offer auto-captioning, so I’ve prepared a script (attached) for you to do it locally on your computer.

For Flux fast and easy Lora training you can still use FluxGym with a clean, fully coded Jupyter Notebook that follows the familiar structure you know and love. Inside the fluxgym/output folder in Jupiter Notebook, you’ll find your ready-to-use LoRA model.

• FluxGym runs on port 8000
• Jupyter Notebook runs on port 8888

You can deploy the template directly here:
https://runpod.io/console/deploy?template=s7y3nvct81&ref=1yinrvfx

This template works only in community cloud.

✍️ New Image Captioning Utility for AI Tools (ostris)Captioning requires matching filenames—e.g. Lara_1.png must have a corresponding Lara_1.txt.
To simplify this, I’ve added a drag-and-drop captioning script:

After unpacking the folder, open PowerShell in the Florence‑2‑Image‑Caption directory and run:

pip install -r requirements.txt

python app.py

You will see a link in powershell terminal how to open your Florence-2-Image-Caption.

This generates a ZIP with captioned images ready for dataset inclusion into AI tools.

🔗 Essential Links:

DruidCat's Youtube: https://www.youtube.com/@AIDruidCat

Runpod:https://get.runpod.io/tupgg9dxnana

FluxGym Tutorial:https://youtu.be/5qvg06ZTHAw
FluxGym Template: https://runpod.io/console/deploy?template=s7y3nvct81&ref=1yinrvfx

Kate LoRA: https://civitai.com/models/1658793/kate-nobody
Ultra Real Fine Tune Flux:https://civitai.com/models/978314/ultrareal-fine-tune

💻 Installation Commands:

Unzip scripts inside ComfyUI folder - path should me like this - ComfyUI/druidcat_scripts/(included folders)

Short instruction - put druidcat_scripts.zip inside ComfyUI folder, right click and choose Extract Archive, go inside scripts, they are modular and you can install what you need. All scripts got around 900GB of space.

Restart and refresh pod after installing scripts.

If you encounter an unauthenticated pull limit request, then please terminate and try again another pod.

Windows: Open PowerShell in the script folder and type: powershell -ExecutionPolicy Bypass -File install.ps1

or Set-ExecutionPolicy Bypass -Scope Process -Force; .\install.ps1

RunPod, Linux, Ubuntu: Run the installation script in terminal: bash install.sh

Why Try It Now?

• Take advantage of RTX 5090 cloud access for maximum generation speed with comfyhub 5090. (Link attached)
• Explore ultra-fast generation using WAN 2.2 as either Text‑to‑Image, Image‑to‑Video, or FLF2V cinematic workflows blog.comfy.org
• Use WAN 2.1 FusionX for very quick, efficient high-res output (both video and image)
• Workflow files are annotated step-by-step to "hold your hand" through the process—bringing ease if you're already familiar with ComfyUI.

🎬 The full free tutorial is still in production. Test WAN 2.2 on RTX 5090, explore WAN 2.1, try Flux Kontext Dev tools — and don’t forget to use the captioning tool for your datasets.

I also attached older Runpod templates from my Youtube channel. https://www.youtube.com/@AIDruidCat

24.08.2025 Update

🔗 Links & Resources

• Stella model: https://civitai.com/models/1886243/st...
• Runpod Ostris AI Tools: https://get.runpod.io/tupgg9dxnana
• Ostris Wan LoRA Trainer (Replicate): https://replicate.com/ostris/wan-lora...
• FLUX.1 Krea model (HuggingFace): https://huggingface.co/black-forest-l...

druidcat_scripts.zip has been updated – most of the previous bugs are now fixed. If you downloaded it earlier and ran into issues, please try again and grab the latest version from this post. ✅

29.09.2025 Update - New versions of the workflows and more

I'm absolutely THRILLED to announce the biggest update yet to our ComfyUI script collection! This isn't just an update - it's a complete revolution in AI art generation. 

Videos showcasing these incredible new features are coming soon!

This script pack is THE ultimate answer for anyone starting with ComfyUI - our automated scripts will install the latest open-source AI models and tools for you, making professional AI art generation accessible to everyone!

Trust me, you NEED to test everything yourself because the results are absolutely mind-blowing!

🚀 What's New in This Extended Update:

🎨 Qwen Text-to-Image: The New September Beast

The Qwen model received a MASSIVE September update that's giving Nano Banana serious competition! The quality improvements are insane - photorealistic results that will make your jaw drop.

📈 Upscalers Revolution: Flux-Powered Realism

Complete game-changer in upscaling! Our new Flux-based upscaler delivers stunningly realistic results with unprecedented detail and clarity.

Plus, we've added the incredible SeedVR2 upscalers for both images AND videos - absolutely ESSENTIAL tools that will become your go-to solution for enhancing any visual content.

⚡ SRPO: The Flux-Based Powerhouse

Meet SRPO - an absolutely POTENT model based on Flux delivering exceptional results as both upscaler AND generator. The quality output is so impressive it's worth testing immediately!

🎬 Wan 2.2 Enhancements: Speed Meets Quality

Supercharged Wan 2.2 workflows with lightning-fast performance improvements. New "fast workflows" cut generation times dramatically while maintaining premium quality. Plus, we squashed numerous bugs from earlier versions.

🎭 Wan 2.2 Vace and Wan Animate: Character Animation Revolution

This is absolutely groundbreaking! The new Wan 2.2 Animate model can animate ANY character based on a performer's video, precisely replicating facial expressions and movements to generate incredibly realistic character videos. It can also replace characters in existing videos while preserving expressions, movements, and seamlessly integrating with original lighting and environments. The dual-mode functionality and advanced body motion control make this a complete game-changer for character animation!

🎭 USO Flux: The Creative Powerhouse

Following Nano Banana and other top-tier editors, our USO Flux pack delivers professional-grade editing with intuitive workflows for beginners and pros alike.

✅ All files are available after purchase.

Thanks so much for your support — it keeps everything evolving!

– Druid Cat 🐾

By purchasing the digital product, you agree to and accept the terms outlined in the document titled "Terms and License Agreement", available at the following link:
https://docs.google.com/document/d/1991V89sdl-tcFidCKpFDOhz9BhSZFlG_UXBRR3cJy2Y/edit?usp=sharing