FlowScale AI - Streamline Collaboration, Simplify Deployment

A collaborative platform for building and deploying AI workflows, enabling teams to create, iterate, and launch generative AI applications. Simplify collaboration, streamline development, and ship workflows with ease.

Responsibilities

Responsibilities

Responsibilities

Responsibilities

Branding, Ideation, User Interviews, User Journey, User Flow, Wireframing, Prototyping, Visual Design, Iterations, Usability testing

Branding, Ideation, User Interviews, User Journey, User Flow, Wireframing, Prototyping, Visual Design, Iterations, Usability testing

Branding, Ideation, User Interviews, User Journey, User Flow, Wireframing, Prototyping, Visual Design, Iterations, Usability testing

Duration

Duration

Duration

Duration

March 2024 - June 2024

March 2024 - June 2024

March 2024 - June 2024

FlowscaleAI Team

FlowscaleAI Team

FlowscaleAI Team

FlowscaleAI Team

1 Project Manager, 1 Product Manager, 2 developers

1 Project Manager, 1 Product Manager, 2 developers

1 Project Manager, 1 Product Manager, 2 developers

FlowScaleAI is a platform for collaborative ComfyUI workflow building and deployment, offering version control, cloud scalability, and seamless environment management. It empowers teams to streamline collaboration, eliminate technical bottlenecks, and scale AI applications effortlessly.

FlowScaleAI is a platform for collaborative ComfyUI workflow building and deployment, offering version control, cloud scalability, and seamless environment management. It empowers teams to streamline collaboration, eliminate technical bottlenecks, and scale AI applications effortlessly.

FlowScaleAI is a platform for collaborative ComfyUI workflow building and deployment, offering version control, cloud scalability, and seamless environment management. It empowers teams to streamline collaboration, eliminate technical bottlenecks, and scale AI applications effortlessly.

FlowScaleAI is a platform for collaborative ComfyUI workflow building and deployment, offering version control, cloud scalability, and seamless environment management. It empowers teams to streamline collaboration, eliminate technical bottlenecks, and scale AI applications effortlessly.

What is ComfyUI?

What is ComfyUI?

What is ComfyUI?

  • ComfyUI is a web-based interface for Stable Diffusion, designed for workflow customization.


  • It uses a modular graph/nodes system to simplify image generation tasks, letting users easily connect nodes for text prompts, image inputs, and AI filters.

  • ComfyUI is a web-based interface for Stable Diffusion, designed for workflow customization.


  • It uses a modular graph/nodes system to simplify image generation tasks, letting users easily connect nodes for text prompts, image inputs, and AI filters.

  • ComfyUI is a web-based interface for Stable Diffusion, designed for workflow customization.


  • It uses a modular graph/nodes system to simplify image generation tasks, letting users easily connect nodes for text prompts, image inputs, and AI filters.

Problems

Problems

Problems

Collaboration Changes

Collaboration Changes

Collaboration Changes

Collaboration Changes

ComfyUI workflow builders face collaboration issues, often sharing JSON files via WhatsApp or Slack without version control. This makes tracking changes hard, causing disruptions and delays due to debugging and conflicts between team members' edits.

ComfyUI workflow builders face collaboration issues, often sharing JSON files via WhatsApp or Slack without version control. This makes tracking changes hard, causing disruptions and delays due to debugging and conflicts between team members' edits.

ComfyUI workflow builders face collaboration issues, often sharing JSON files via WhatsApp or Slack without version control. This makes tracking changes hard, causing disruptions and delays due to debugging and conflicts between team members' edits.

ComfyUI workflow builders face collaboration issues, often sharing JSON files via WhatsApp or Slack without version control. This makes tracking changes hard, causing disruptions and delays due to debugging and conflicts between team members' edits.

Limited Local Resources

Limited Local Resources

Limited Local Resources

Limited Local Resources

Heavy workflows, like video processing, overwhelm local machines. Teams must set up cloud infrastructure, adding complexity and making tasks like generating a 3-minute video inefficient and time-consuming.


Heavy workflows, like video processing, overwhelm local machines. Teams must set up cloud infrastructure, adding complexity and making tasks like generating a 3-minute video inefficient and time-consuming.


Heavy workflows, like video processing, overwhelm local machines. Teams must set up cloud infrastructure, adding complexity and making tasks like generating a 3-minute video inefficient and time-consuming.


Heavy workflows, like video processing, overwhelm local machines. Teams must set up cloud infrastructure, adding complexity and making tasks like generating a 3-minute video inefficient and time-consuming.


Dependency and Environment Issues

Dependency and Environment Issues

Dependency and Environment Issues

Dependency and Environment Issues

Dependency conflicts, especially with Python, cause workflows that work on one machine to fail on another, wasting time troubleshooting environment-specific problems.

Dependency conflicts, especially with Python, cause workflows that work on one machine to fail on another, wasting time troubleshooting environment-specific problems.

Dependency conflicts, especially with Python, cause workflows that work on one machine to fail on another, wasting time troubleshooting environment-specific problems.

Dependency conflicts, especially with Python, cause workflows that work on one machine to fail on another, wasting time troubleshooting environment-specific problems.

Dependency conflicts, especially with Python, cause workflows that work on one machine to fail on another, wasting time troubleshooting environment-specific problems.

Lack of Accessible Deployment Tools

Lack of Accessible Deployment Tools

Lack of Accessible Deployment Tools

Lack of Accessible Deployment Tools

Non-developers, like designers, rely on developers to deploy workflows, slowing iteration. There’s no easy, one-click tool for them to deploy and share workflows independently.

Non-developers, like designers, rely on developers to deploy workflows, slowing iteration. There’s no easy, one-click tool for them to deploy and share workflows independently.

Non-developers, like designers, rely on developers to deploy workflows, slowing iteration. There’s no easy, one-click tool for them to deploy and share workflows independently.

Non-developers, like designers, rely on developers to deploy workflows, slowing iteration. There’s no easy, one-click tool for them to deploy and share workflows independently.

Non-developers, like designers, rely on developers to deploy workflows, slowing iteration. There’s no easy, one-click tool for them to deploy and share workflows independently.

Goal

Goal

Goal

  • Develop a centralized platform for real-time collaboration, with version control and change tracking, reducing reliance on external tools for workflow sharing.


  • Integrate automatic versioning, rollback, and visual change logs for efficient tracking and troubleshooting.


  • Optimize scalability by integrating cloud GPU infrastructure and automating resource management based on workflow demands.


  • Simplify deployment for non-developers with one-click deployment and drag-and-drop interfaces, providing pre-built templates for scalability.

  • Develop a centralized platform for real-time collaboration, with version control and change tracking, reducing reliance on external tools for workflow sharing.


  • Integrate automatic versioning, rollback, and visual change logs for efficient tracking and troubleshooting.


  • Optimize scalability by integrating cloud GPU infrastructure and automating resource management based on workflow demands.


  • Simplify deployment for non-developers with one-click deployment and drag-and-drop interfaces, providing pre-built templates for scalability.

  • Develop a centralized platform for real-time collaboration, with version control and change tracking, reducing reliance on external tools for workflow sharing.


  • Integrate automatic versioning, rollback, and visual change logs for efficient tracking and troubleshooting.


  • Optimize scalability by integrating cloud GPU infrastructure and automating resource management based on workflow demands.


  • Simplify deployment for non-developers with one-click deployment and drag-and-drop interfaces, providing pre-built templates for scalability.

Solution

Solution

Solution

Seamlessly version-control your ComfyUI workflows

Seamlessly version-control your ComfyUI workflows

Seamlessly version-control your ComfyUI workflows

  • Users can make changes in their local system and save it to the repository on the cloud.


  • The repositories are connected to Github account where users can view the code for each file.


  • Users can upload new folders or files directly to the repository from the Flowscale dashboard.

Collaborate on development of ComfyUI workflows on remote repositories.

Collaborate on development of ComfyUI workflows on remote repositories.

Collaborate on development of ComfyUI workflows on remote repositories.

  • Users can save their changes from the local system to the cloud by committing and synchronising their changes.


  • Users can also view the Change Log which is a history of all the changes made by the team.


  • This allows users to collaborate real-time and reduces the confusion and delay.

  • Users can save their changes from the local system to the cloud by committing and synchronising their changes.


  • Users can also view the Change Log which is a history of all the changes made by the team.


  • This allows users to collaborate real-time and reduces the confusion and delay.

  • Users can save their changes from the local system to the cloud by committing and synchronising their changes.


  • Users can also view the Change Log which is a history of all the changes made by the team.


  • This allows users to collaborate real-time and reduces the confusion and delay.

Build deployment pipelines to take your ComfyUI workflows to production.

Build deployment pipelines to take your ComfyUI workflows to production.

Build deployment pipelines to take your ComfyUI workflows to production.

  • Users can deploy their workflow by publishing the workflow.


  • After deploying, users will be able to see their deployments.


  • They also get a ready-to-go UI which they can share with other users to test the workflow.


  • Users also get the API documents for the workflow which they can use to embed this workflow into others.

  • Schedule tasks and workflows for individuals or teams.


  • Use ready templates or create your own to schedule an insight.

  • Schedule tasks and workflows for individuals or teams.


  • Use ready templates or create your own to schedule an insight.

User Research

User Research

User Research

User Research

We conducted qualitative research with six participants: three developers and three non-developers (including 3D designers and motion designers). All participants were professionals from the technical field, providing insights from both development and creative perspectives.

We conducted qualitative research with six participants: three developers and three non-developers (including 3D designers and motion designers). All participants were professionals from the technical field, providing insights from both development and creative perspectives.

Some of the questions we asked the users:

Some of the questions we asked the users:

How did you get started with Comfy UI?

What did you use ComfyUI for?

How did you collaborate with your team?

Is it difficult to collaborate with your team?

What are the difficulties faced when sharing the workflows?

Have you deployed anything you created on ComfyUI to scale? Why not?

How did you get started with Comfy UI?

What did you use ComfyUI for?

How did you collaborate with your team?

Is it difficult to collaborate with your team?

What are the difficulties faced when sharing the workflows?

Have you deployed anything you created on ComfyUI to scale? Why not?

How did you get started with Comfy UI?

What did you use ComfyUI for?

How did you collaborate with your team?

Is it difficult to collaborate with your team?

What are the difficulties faced when sharing the workflows?

Have you deployed anything you created on ComfyUI to scale? Why not?

How did you get started with Comfy UI?

What did you use ComfyUI for?

How did you collaborate with your team?

Is it difficult to collaborate with your team?

What are the difficulties faced when sharing the workflows?

Have you deployed anything you created on ComfyUI to scale? Why not?

Key insights derived from the interviews

Key insights derived from the interviews

Familiarity with Node-Based Systems

Familiarity with Node-Based Systems

Familiarity with Node-Based Systems

  • Designers with experience in tools like Blender adapted easily.

  • Non-developers struggled to keep up with the system.

Limited Learning Resources

Limited Learning Resources

Limited Learning Resources

  • A lack of tutorials and documentation made it hard for non-developers to fully utilize the platform.

Deployment Challenges for Non-Developers

Deployment Challenges for Non-Developers

Deployment Challenges for Non-Developers

  • Non-developers had difficulty deploying without coding knowledge, often relying on developers for help.

Manual Collaboration Processes

Manual Collaboration Processes

Manual Collaboration Processes

  • Using WhatsApp, Notion, and Loom for collaboration was slow and repetitive.

  • An integrated workflow management tool is needed within ComfyUI.

Technical Stability Issues

Technical Stability Issues

Technical Stability Issues

  • Frequent node import failures and system instability on Windows caused frustration for daily users.

Who we designed for

Who we designed for

Who we designed for

Who we designed for

Software Developers

Software Developers

Software Developers

Software Developers

Users who can write code and can build ComfyUI workflows

Users who can write code and can build ComfyUI workflows

Users who can write code and can build ComfyUI workflows

Users who can write code and can build ComfyUI workflows

Non-Technical Users

Non-Technical Users

Non-Technical Users

Non-Technical Users

Users who don’t know to code but can use ComfyUI

Users who don’t know to code but can use ComfyUI

Users who don’t know to code but can use ComfyUI

Users who don’t know to code but can use ComfyUI

Current User Journey

Current User Journey

Current User Journey

Current User Journey

Proposed User Journey

Proposed User Journey

Proposed User Journey

Proposed User Journey

Branding

Branding

Branding

Branding

We decided to go with a blue-green gradient color as the brand color as the combination looks futuristic and are trending as AI colors.

Blue is a color often used in the logos of AI companies to convey authority, trust, and dependability and green is a growing color choice for most AI products.

We decided to go with a blue-green gradient color as the brand color as the combination looks futuristic and are trending as AI colors.

Blue is a color often used in the logos of AI companies to convey authority, trust, and dependability and green is a growing color choice for most AI products.

We decided to go with a blue-green gradient color as the brand color as the combination looks futuristic and are trending as AI colors.

Blue is a color often used in the logos of AI companies to convey authority, trust, and dependability and green is a growing color choice for most AI products.

We decided to go with a blue-green gradient color as the brand color as the combination looks futuristic and are trending as AI colors.

Blue is a color often used in the logos of AI companies to convey authority, trust, and dependability and green is a growing color choice for most AI products.

flowscale.ai

flowscale.ai

flowscale.ai

flowscale.ai

flowscale.ai

flowscale.ai

First Version of the Product

First Version of the Product

First Version of the Product

First Version of the Product

We identified the need for two key components in the product:


  1. Local Application: A tool for users to install on their machines, allowing them to save and manage changes locally.

  2. Collaborative Dashboard: A centralized dashboard where users can access workflows, track team members' updates, and view deployment progress in real-time.

We identified the need for two key components in the product:


  1. Local Application: A tool for users to install on their machines, allowing them to save and manage changes locally.

  2. Collaborative Dashboard: A centralized dashboard where users can access workflows, track team members' updates, and view deployment progress in real-time.

We identified the need for two key components in the product:


  1. Local Application: A tool for users to install on their machines, allowing them to save and manage changes locally.

  2. Collaborative Dashboard: A centralized dashboard where users can access workflows, track team members' updates, and view deployment progress in real-time.

We identified the need for two key components in the product:


  1. Local Application: A tool for users to install on their machines, allowing them to save and manage changes locally.

  2. Collaborative Dashboard: A centralized dashboard where users can access workflows, track team members' updates, and view deployment progress in real-time.

We initially began with high-fidelity designs for the nodes due to time constraints and because our brand guide was already prepared. Our goal was to quickly test the product with users we had interviewed to ensure we were heading in the right direction. The features we focused on included local setup, file structure and commit changes.

We initially began with high-fidelity designs for the nodes due to time constraints and because our brand guide was already prepared. Our goal was to quickly test the product with users we had interviewed to ensure we were heading in the right direction. The features we focused on included local setup, file structure and commit changes.

We initially began with high-fidelity designs for the nodes due to time constraints and because our brand guide was already prepared. Our goal was to quickly test the product with users we had interviewed to ensure we were heading in the right direction. The features we focused on included local setup, file structure and commit changes.

We initially began with high-fidelity designs for the nodes due to time constraints and because our brand guide was already prepared. Our goal was to quickly test the product with users we had interviewed to ensure we were heading in the right direction. The features we focused on included local setup, file structure and commit changes.

Usability Testing

Usability Testing

Usability Testing

Usability Testing

Usability Testing

We tested the product with five users and observed several areas of confusion and frustration:

We tested the product with five users and observed several areas of confusion and frustration:

We tested the product with five users and observed several areas of confusion and frustration:

Node Design Confusion


Users were unaware that a dashboard was also being developed, leading to confusion with the node design.

File Management Issues


Users were frustrated by the limitation of opening only one file at a time and the lack of state maintenance.

Branch Confusion


Users were unclear about the purpose of the "dev-mode" feature located at the bottom.

Commit Textbox Misunderstanding


The textbox for commits was only understood by developers, as it resembled the Visual Studio Code (VSCode) user interface.

Deployment Picture Incompleteness


Users struggled to form a complete understanding of how to deploy the ComfyUI workflow they created.

UX Brainstorming for the Local Installation

UX Brainstorming for the Local Installation

UX Brainstorming for the Local Installation

UX Brainstorming for the Local Installation

Each feature was thought through with wireframing before jumping into the next version of the product so it meets the user’s requirements.

Each feature was thought through with wireframing before jumping into the next version of the product so it meets the user’s requirements.

Each feature was thought through with wireframing before jumping into the next version of the product so it meets the user’s requirements.

UX Brainstorming for the Dashboard

UX Brainstorming for the Dashboard

UX Brainstorming for the Dashboard

UX Brainstorming for the Dashboard

The initial plan for the dashboard was to have a ComfyUI view embedded in the dashboard and allow users to collaborate in real time (like Figma). However, this was put on hold due to technical and time constraints.

The initial plan for the dashboard was to have a ComfyUI view embedded in the dashboard and allow users to collaborate in real time (like Figma). However, this was put on hold due to technical and time constraints.

The initial plan for the dashboard was to have a ComfyUI view embedded in the dashboard and allow users to collaborate in real time (like Figma). However, this was put on hold due to technical and time constraints.

We went ahead with the folder structure, deployments, runs, and settings pages. Workspaces was added eventually too.

We went ahead with the folder structure, deployments, runs, and settings pages. Workspaces was added eventually too.

We went ahead with the folder structure, deployments, runs, and settings pages. Workspaces was added eventually too.

Final Product

Final Product

Final Product

Final Product

Thank you for reading!

Thank you for reading!

Thank you for reading!

Thank you for reading!

Go back to top

Go back to top

Go back to top

Let's

craft

something

extraordinary! 🤝

Let's

craft

something

extraordinary! 🤝

Let's

craft

something

extraordinary! 🤝

Let's

craft

something

extraordinary! 🤝

Let's

craft

something

extraordinary! 🤝