Studio5090
A film studio pipeline for ComfyUI, designed by a 40-year VFX artist. Requires the Free STUDIO5090 starter project. Studio5090 offers to install it on first start. Cast it, create it, render it. CASTING builds a character with a voice and an accent and auditions them on camera. CREATE makes images, character sheets and full multi-angle sets with FLUX.2 and QWEN. RENDER turns them into finished video with LTX-2.3, first frame to last. Demos at vimeo.com/showcase/12492601.
Studio5090
Studio5090 is a series of ComfyUI workflows designed by a VFX artist, built the way a film studio works: cast it, create it, render it.
Designed by Alan Rosenfeld.
What you need
Studio5090 runs inside ComfyUI. If you don't have ComfyUI yet, install it first:
- Download ComfyUI from the official site: comfy.org/download. ComfyUI Desktop is the easiest way to start on Windows and macOS.
- Installation guides for every version (Desktop, Portable and manual) are at docs.comfy.org.
- ComfyUI Manager is used to install the other node packs the workflows need. ComfyUI Desktop includes it. If your ComfyUI doesn't have it, see the ComfyUI Manager page.
- An NVIDIA graphics card. Studio5090 is developed on an RTX 5090 and tested on an RTX 3070. On smaller cards, choose a smaller Prompt Generator model (see STEP 2 below).
Start ComfyUI once and make sure it runs before installing Studio5090.
Where your ComfyUI folders are: the steps below refer to ComfyUI/output/ and ComfyUI/custom_nodes/. In ComfyUI Desktop, these folders are inside the location you chose when you installed it.
Before you begin: download the starter project
Studio5090 needs its starter project to be in place before you open any workflow. The workflows, examples and test files all live inside it.
Go to the Releases page and download these three files:
| File | What it is |
|---|---|
| STUDIO5090.zip | The starter project. Required. |
| Studio5090_Documentation.pdf | How every part of Studio5090 works. |
| Studio5090_Bible.pdf | Recipes, render style samples and tricks. |
Installation
1. Extract the starter project
Extract STUDIO5090.zip into your ComfyUI output folder.
On Windows: right-click STUDIO5090.zip and choose Extract All. Windows adds \STUDIO5090 to the end of the destination path. Delete that part, so the path ends in \ComfyUI\output, then click Extract.
When you are done, this folder must exist:
ComfyUI/output/STUDIO5090/STUDIO5090/
If the folder ends up one level too deep (STUDIO5090/STUDIO5090/STUDIO5090), Studio5090 tells you when ComfyUI starts. Move everything inside the inner STUDIO5090 folder up one level, then delete the empty inner folder.
2. Install the Studio5090 nodes
With git: open a terminal in ComfyUI/custom_nodes/ and run
git clone https://github.com/Studio5090/ComfyUI-Studio5090-Utils
Without git: on this page, click the green Code button, then Download ZIP. Unzip it, rename the folder from ComfyUI-Studio5090-Utils-main to ComfyUI-Studio5090-Utils, and move it into ComfyUI/custom_nodes/.
Keep only one copy of Studio5090 in custom_nodes. If you had an earlier copy, remove it first.
3. Restart ComfyUI
Restart ComfyUI completely, not just the browser page.
First run
STEP 1: Open the Casting template
Open the Casting template first. It contains every model Studio5090 uses, so it is the best way to get everything installed at once.
The templates are in the starter project you extracted: ComfyUI/output/STUDIO5090/STUDIO5090/workflows/
Drag CASTING_TEMPLATE.json onto ComfyUI to open it. This folder also holds the example project workflows. The same templates are in ComfyUI/custom_nodes/ComfyUI-Studio5090-Utils/templates/product_workflows/.
- Use ComfyUI Manager > Install Missing Custom Nodes to install the other node packs the workflow needs, then restart ComfyUI.
- Download any missing models. ComfyUI lists them when the workflow opens.
MODELS.mdin this repository lists every model and the folder it goes in.
STEP 2: Set the Prompt Generator model
Every time you open a template, after the models and nodes are installed, set the Prompt Generator model first. The Documentation and the demo videos show where to do this.
Studio5090's LLM Model Select node sets the model for every Prompt Generator in the workflow at once, so you only pick it in one place.
STEP 3: Run Casting
Run the Casting workflow.
If you get errors, the most likely cause is a Prompt Generator model that is too big for your system. Choose a smaller model in LLM Model Select and run it again.
The three tools
Casting: create a character, give them a voice and an accent, and audition them on camera.
Create: build images and character sheets with FLUX.2 and QWEN, including full sets of multi-angle views.
RENDER: turn your images and audio into finished video with LTX-2.3, from first frame to last frame.
Watch each one in action in the Studio5090 Vimeo Showcase.
Your own copies of the text files
Accents, render styles and the LLM instructions are plain text files. To change one, copy it into ComfyUI/output/STUDIO5090/ and edit the copy. Your copy is used in place of the toolkit's, and updates never overwrite it. See STUDIO5090_TIPS.md.
License
MIT License. See LICENSE and THIRD_PARTY_LICENSES.md.