Save 4 hours! We auto-setup your workflow! Free!

Drop your workflow.json — we handle every dependency, custom node, and model. Just open the link and run.

Auto-Setup Workflow Json (Free) Now!
ComfyUI > Nodes > FunUI

ComfyUI Extension: FunUI

Repo Name

FunUI

Author
digital-garbage (Account age: 1276 days)
Nodes
View all nodes(20)
Latest Updated
2026-11-03
Github Stars
0.02K

How to Install FunUI

Install this extension via the ComfyUI Manager by searching for FunUI
  • 1. Click the Manager button in the main menu
  • 2. Select Custom Nodes Manager button
  • 3. Enter FunUI in the search bar
After installation, click the Restart button to restart ComfyUI. Then, manually refresh your browser to clear the cache and access the updated list of nodes.

Visit ComfyUI Online for ready-to-use ComfyUI environment

  • Free trial available
  • 16GB VRAM to 80GB VRAM GPU machines
  • 400+ preloaded models/nodes
  • Freedom to upload custom models/nodes
  • 200+ ready-to-run workflows
  • 100% private workspace with up to 200GB storage
  • Dedicated Support

Run ComfyUI Online

FunUI Description

FunUI is a video-production workspace inside ComfyUI, combining scene prompts, model pipeline settings, clip review, timeline organization, montage tools, and optional LoRA training modules.

FunUI Introduction

FunUI is a browser-based video studio inside ComfyUI. It brings story writing, scene generation, clip review, and timeline organization into one workspace, while using the regular ComfyUI backend to run your models.

Instead of starting with a large graph of connected nodes, you can work with tools that are closer to a creative production workflow:

  • A timeline with one clip per scene to organize your story.
  • Composer to write multiple scene prompts together.
  • Pipeline cards to choose model files and configure the generation workflow.
  • Takes and ratings to review results and help supported modules adapt to your preferences.
  • Auto Montage to create a new cut from clips you have already rendered.
  • Optional modules that you can enable or disable.
  • A LoRA training add-on to teach a model from your own media and rated results. FunUI can help if you want to spend less time navigating node graphs and more time developing scenes, comparing variations, and shaping a sequence.

FunUI v5 is a complete rewrite, not an incremental update to version 4.x. You access its studio at /fun/ on your running ComfyUI server.

FunUI studio showing the preview, scene settings, and a six-clip timeline

About the screenshots: The images in this README show the real application using an abstract demonstration project. Its clips were made with ffmpeg, not generated by an AI model.

Why the name changed

FunUI was formerly called FunPack. The project began as two nodes for experimenting with the FramePack video model. As it grew into a dedicated interface on top of ComfyUI, the author renamed it to better describe what it had become: a complete UI rather than simply a node pack.

FunUI is a free, open-source hobby project maintained by the author. Testing resources are limited, and some features remain experimental. See CHANGELOG.md for the documented distinction between proven and experimental functionality.

How FunUI Works

Think of FunUI as the studio desk and ComfyUI as the engine room. You arrange scenes and make creative decisions in FunUI; ComfyUI performs the underlying model processing.

1. Set up a project

The Project Setup Wizard guides you through choosing a look, naming your project, and loading the pipeline for the model you intend to use.

A pipeline is the sequence of processing steps needed to turn your prompts and references into an output. Different models need different pipelines.

2. Turn your story into scenes

Composer keeps your story in one text box. A chosen cut word marks where one scene ends and another begins.

For example, you could plan:

  • An opening view of a rain-soaked street.
  • A closer view of a character under a streetlamp.
  • A final view of the character walking away. The cut word separates these descriptions into scene prompts. Reusable variables, templates, and shortcuts reduce repeated typing.

3. Configure the generation process

In Models & Pipeline, the workflow appears as cards rather than a node graph.

Each card represents part of the process, such as loading a model or passing information to another step. You choose the required files and connect the appropriate outputs and inputs.

FunUI checks whether the pipeline can be built. If something required is missing or invalid, the server rejects it and explains why. This helps you correct the setup before attempting generation.

4. Generate and review your clips

Each scene has a clip on the timeline. The studio includes a preview, takes, ratings, and Render and Export controls, keeping generation and review in the same workspace.

Quick sampler controls let you adjust common generation settings without leaving the clip you are working on.

5. Use ratings to guide supported modules

When you rate clips, the guidance, steering, and prompt modules can begin adapting to your preferences after enough ratings have been collected.

Think of this as giving creative feedback: the modules need examples of what you like before they can respond to your taste.

This is separate from LoRA training. Ratings-based adaptation and training a new LoRA are different features.

6. Keep optional features modular

In v5, features register themselves as modules and expose their settings to the interface.

If an optional module fails to load, FunUI can leave it out and continue the run, with a notice. This does not mean a pipeline can run without its required model files or essential nodes.

FunUI Features

Timeline studio

The timeline organizes your project with one clip per scene. The surrounding workspace includes:

| Tool | What it helps you do | |---|---| | Media bin | Keep project media accessible in the studio. | | Preview | Inspect the clip you are working on. | | Takes | Work with alternative results for a scene. | | Ratings | Record your preferences and provide feedback to learning-enabled modules. | | Render and Export | Access output controls from the project workspace. | | Guided tour | Learn the interface through an in-app walkthrough. |

This layout is useful when you are developing a sequence rather than treating each generation as an isolated result.

Composer: write the story once

Composer combines several prompt-writing tools:

  • Cut word: Divide one story text into separate scenes.
  • $variables: Reuse descriptions that appear throughout your project.
  • Templates: Keep reusable prompt structures close at hand.
  • Shortcuts: Reduce repetitive text entry.
  • Optional Enhance tab: Use a language model to polish each prompt. For example, a reusable variable could hold a character description or a recurring visual treatment. You can then reference that wording across scene prompts instead of typing it repeatedly.

Variables help you keep descriptions consistent; they do not, by themselves, guarantee visual consistency in generated results.

FunUI Composer Story tab

The Enhance tab is optional. It is useful when you want help expanding or refining wording, but you should review the revised prompts to make sure they still express your intention.

FunUI Composer Enhance tab

Project Setup Wizard

The wizard presents project setup one step at a time:

  1. Choose your look.
  2. Name the project.
  3. Load the pipeline for your chosen model. This provides a more approachable starting point than navigating multiple settings tabs before you have begun writing.

FunUI Project Setup Wizard

Models & Pipeline: cards instead of graphs

The card-based pipeline view lets you:

  • Select the files required by each card.
  • Swap or add nodes, including installed custom nodes.
  • Connect outputs to inputs.
  • Import a ComfyUI workflow.
  • See whether required slots are filled.
  • Identify the first missing requirement when the pipeline is incomplete. For an artist, the main benefit is visibility: you can see what the workflow needs without beginning with a large graph.

Adding an installed node does not automatically make it compatible with every pipeline. FunUI still validates whether the resulting workflow can be built.

FunUI Models & Pipeline cards

Modules and Engine settings

Use Settings → Modules to turn individual modules on or off.

Modules announce their own settings, so loaded modules can expose their controls in Engine settings rather than relying on a separately maintained list.

You can use this to keep optional features out of the way while concentrating on a simpler workflow.

FunUI Engine settings

FunUI module switches

Quick sampler settings

The quick settings panel provides access to steps, sampler, scheduler, seed, and common switches from the timeline.

| Setting | Plain-language explanation | Practical effect | |---|---|---| | Steps | How many refinement passes generation uses. | More steps generally require more processing time, but do not always produce a better result. Follow the model or preset’s intended settings. | | Sampler | The method used to refine the generated result. | Changing it can alter how a result develops. | | Scheduler | How refinement is distributed across the steps. | Its effect depends on the model and sampler combination. | | Seed | A number used to establish the starting randomness. | Keeping it fixed can help you compare setting changes; changing it can explore a different variation. |

A fixed seed is helpful for comparisons, but it does not guarantee identical output when you change models, workflows, or other settings.

Few-step methods such as PTD require their matching adapters and sampling setup. Simply reducing the step count is not equivalent.

FunUI quick sampler settings

Customizable picker wheel

Press the middle mouse button to open a wheel of frequently used actions under your pointer. Release over an action to run it.

You choose which actions appear, allowing you to keep your most-used commands nearby.

FunUI picker wheel

Learning from ratings

The guidance, steering, and prompt modules can use your clip ratings once they have enough examples.

This gives your review process a second purpose: besides helping you judge a take, your feedback can influence supported modules’ behavior.

The provided documentation does not specify a universal minimum rating count or promise a particular improvement. Treat this as an adaptive feature, not a guarantee that every subsequent clip will match your preferences.

Auto Montage

Auto Montage builds a new cut at the end of the timeline using your rendered clips.

Your original clips remain unchanged, so you can inspect the new cut without replacing the material you started with.

FunUI Auto Montage

LoRA training add-on

A LoRA is a smaller set of learned adjustments used alongside a base model. It can help teach that model a particular style or concept.

FunUI’s training add-on uses your own media and results you have rated. Its basic workflow is:

  1. Enable the add-on in Settings → Modules.
  2. Select your training data.
  3. Describe each item.
  4. Press Start. The add-on has only been run on MiniMax H3. Do not assume that support for generating with another model also means its training is supported.

FunUI LoRA training data selection

FunUI LoRA training descriptions and tags

FunUI LoRA training controls

The implementation follows the MiniMax H3 training recipe from musubi-tuner. Musubi Tuner’s broader model support and settings should not be assumed to be available in FunUI.

Dark and light themes

FunUI includes dark and light appearances, so you can choose the workspace that is more comfortable for your viewing environment.

FunUI studio in the light theme

FunUI Models

Model support is not equally mature across all options.

| Model | Documented status | When to consider it | |---|---|---| | MiniMax H3 | The model v5 is built and tested around. | Start here if you want the model at the center of FunUI’s development and testing, or want to explore the training add-on. | | Stable Diffusion 1.5 | Experimental. | Use for experimentation with that model, with the understanding that its FunUI integration is not presented as fully proven. | | LTX 2.3 | Tested with a real model for the documented image-to-video and video-to-video paths. | Use when you want an LTX workflow driven by an image reference, a source video, or text when no input is supplied. |

MiniMax H3

MiniMax H3 is FunUI v5’s primary model.

The author also documents PTD few-step support for H3, ported from ComfyUI-PTD-Compact.

PTD uses specially trained adapters to generate with fewer sampling steps. The upstream compact project documents:

  • EPFL fl2va adapters for 4 or 6 steps.
  • Ref2VA adapters for 4 or 8 steps.
  • A RefAsymOPD Ref2VA adapter for 8 steps. These are upstream adapter options, not a confirmation that every variant has been validated in FunUI. Check FunUI’s changelog and the relevant model or adapter documentation before choosing a combination. Adapters must match their intended base model and sampling setup.

For background, see the PTD project page and official PTD implementation.

Stable Diffusion 1.5

Stable Diffusion 1.5 support is marked experimental.

The provided documentation does not establish detailed supported workflows or performance expectations for this integration. Use it with that limitation in mind rather than expecting the same coverage as MiniMax H3. ### LTX 2.3 FunUI documents two separate LTX presets:

  • Image to Video: Starts from an image selected in the scene’s References selector.
  • Video to Video: Uses the direct video-to-video path. Both fall back to text-to-video when no input is supplied.

Important workflow details:

  • Choose images in the scene’s References selector; dropping pictures onto the timeline is no longer the LTX input workflow.
  • Switch the preset to change between Image to Video and Video to Video.
  • The integration uses separate model files.
  • LoRAs and Noise merging are documented as supported.
  • Noise merging and the Attention explorer are the enhancements offered on LTX. Other enhancements are held back until proven. The upstream ComfyUI-LTXVideo project includes many additional workflows and features. Their availability upstream does not establish that they are exposed or validated in FunUI.

See the LTX 2.3 model page for model-specific information.

What's New with FunUI

FunUI v5: a complete rewrite

The documented v5 changes include:

| Change | Why it matters | |---|---| | Timeline-based studio | Keeps scene organization, preview, takes, ratings, and output controls together. | | Composer | Lets you write a whole story in one place and reuse wording across scenes. | | Card-based pipelines | Makes model-file selection and workflow requirements easier to inspect. | | Server-side pipeline validation | Explains why a workflow cannot be built before it runs. | | Modular features and settings | Lets you switch optional functionality on or off and see loaded modules’ controls. | | Optional-module failure handling | Allows a run to continue without a missing optional module, with a notice. | | Learning from ratings | Allows supported guidance, steering, and prompt modules to respond to your preferences. | | Auto Montage | Creates another cut while preserving your original clips. | | Project wizard and guided tour | Provides a more approachable introduction to the workspace. |

The current README also documents quick sampler controls, a customizable picker wheel, light and dark themes, the H3 training add-on, and H3 PTD support. Consult CHANGELOG.md for their detailed history and validation status.

LTX workflow changes

LTX image inputs now belong in the scene’s References selector. Image to Video and Video to Video are separate presets, so changing the input media alone is not the documented way to switch workflows.

Classic FunUI 4.x

Version 4 was a different application called Cutting Room, with Easy Gen as its simpler mode and no module system.

It is no longer developed, but remains available on the v4-classic branch.

Do not assume that instructions for 4.x describe v5. You can check your version in Settings → About FunUI.

Troubleshooting FunUI

“I only see the regular ComfyUI interface”

  1. Open your running ComfyUI server in the browser.
  2. Navigate to /fun/ on that same server.
  3. Once FunUI opens, select Help → Start tour if you need an introduction. FunUI is a dedicated workspace, not simply a collection of nodes you must operate from the normal graph view.

“The pipeline will not run”

  1. Open Models & Pipeline.
  2. Read the validation message.
  3. Find the first missing or invalid requirement FunUI identifies.
  4. Select the required file or correct the relevant connection.
  5. Repeat until the pipeline is complete. For imported workflows, remember that essential custom nodes must be available to ComfyUI. Optional-module fallback does not replace required workflow components.

Use the pipeline screenshot as a visual reference.

“A feature or module is missing”

  1. Check Settings → Modules to see whether it is enabled.
  2. Check whether it is supported for your selected model.
  3. Look for a notice that an optional module was left out.
  4. Consult CHANGELOG.md for its status. On LTX, the available enhancements are deliberately limited to Noise merging and the Attention explorer until others are proven.

“LTX ignores my image or generates from text instead”

  1. Select the Image to Video preset.
  2. Open the scene’s References selector.
  3. Choose the intended image there.
  4. Do not rely on dropping the image onto the timeline.
  5. Try the scene again. The documented fallback is text-to-video when no input is supplied.

For a source video, switch to the separate Video to Video preset.

“My ratings do not seem to change anything”

  1. Confirm that the relevant guidance, steering, or prompt module is enabled.
  2. Continue rating clips; adaptation starts only after enough ratings exist.
  3. Do not expect one rating to trigger an immediate visible change.
  4. Keep this feature separate from LoRA training, which requires its own data-selection and training steps.

“The LoRA training tools are not visible”

  1. Open Settings → Modules.
  2. Enable the training add-on.
  3. Use MiniMax H3, the only model on which the add-on is documented as having been run.
  4. Follow the data, descriptions, and training stages shown in the training screenshots.

“A .gguf encoder cannot be read”

FunUI relies on ComfyUI-GGUF to read quantized .gguf encoders when that extension is available.

  1. Check whether the selected encoder file uses .gguf.
  2. Confirm that ComfyUI-GGUF is available and loading in your ComfyUI environment.
  3. Check that the pipeline’s encoder loader matches the file type.
  4. Review the upstream documentation if loading still fails. GGUF support does not mean every GGUF model is compatible with every FunUI pipeline.

Frequently asked questions

Do I need to understand node graphs? You can begin with the project wizard and pipeline cards. Understanding connections becomes more useful when importing or customizing workflows.

Does Auto Montage replace my clips? No. It creates a new cut at the end of the timeline and leaves the originals unchanged.

Is rating a clip the same as training a LoRA? No. Ratings can influence supported adaptive modules. LoRA training is a separate add-on workflow.

Are all supported models equally tested? No. MiniMax H3 is the main development target, Stable Diffusion 1.5 is experimental, and LTX has specific tested paths and deliberately limited enhancements.

Do all upstream LTX or PTD features work in FunUI? Do not assume so. Use FunUI’s documentation and changelog to establish its supported scope.

Where can I report an unresolved problem? Use the FunUI issue tracker. Include your FunUI version, selected model and preset, steps to reproduce, exact error messages, and a screenshot where useful. Avoid including private media or sensitive information.

Learn More about FunUI

Start with the artist-facing guides

Explore model-specific resources

  • LTX open-source ComfyUI documentation — background on LTX workflows and model use.
  • ComfyUI-LTXVideo — upstream LTX nodes and example workflows.
  • PTD project page — an overview of few-step video generation.
  • ComfyUI-PTD-Compact — adapter compatibility and sampling details for the implementation behind FunUI’s H3 PTD support.

Learn about LoRA training

FunUI License and Responsible Use

FunUI is provided under the GNU General Public License v3. You may use, modify, and redistribute the software subject to that license’s terms.

Model weights, adapters, and third-party components may have separate licenses. FunUI’s software license does not replace those terms or determine every right associated with generated content.

The author does not endorse, and will not provide support for, uses involving non-consensual explicit depictions of real people, explicit depictions of minors, or violence and gore targeted at real people.

Use your own media and reference material responsibly, respect consent and applicable rights, and follow your local laws.

FunUI Related Nodes

RunComfy
Copyright 2025 RunComfy. All Rights Reserved.

RunComfy is the premier ComfyUI platform, offering ComfyUI online environment and services, along with ComfyUI workflows featuring stunning visuals. RunComfy also provides AI Models, enabling artists to harness the latest AI tools to create incredible art.

FunUI detailed guide | ComfyUI