This guide is designed to help TD-Teams and TD-EDU users deliver the best experience for sessions, classes, and training events. It covers onboarding, key links, do’s and don’ts, setting up workflows, and common issues to ensure smooth operation.

For TD-EDU

For TD-Teams

Ensuring Sufficient Funds for Large Classes

If you’re planning to run a large class with, say, 50 students, make sure to add sufficient funds to your team account so that all 50 students have enough credits to launch machines without interruptions.

Here’s why this is important:

  • When a student launches a machine, the system pre-debits the funds from your total team balance to secure enough resources for the machine's runtime.
  • If your team balance doesn’t have enough funds, students will face issues launching their machines — even if they haven’t consumed their full usage yet.
  • For example, if each student requires $10 worth of credits to launch a machine, you’ll need at least $500 pre-loaded to ensure everyone can start simultaneously.

To avoid this, you can:

  1. Pre-load sufficient funds in your account before class starts.
  2. Monitor usage under Manage Teams > Set Members Limit to ensure fair usage.
  3. If a student runs into credit issues, you can manually increase their limit or add more funds during the class.

Setup Credit Spend for Each Student:

To prevent abuse or accidental overuse, you can set individual monthly usage limits for each student or workspace member.

If this setting is OFF, users will have unrestricted access to the total credits available in your account.

To set limits:

In your Team Members section, click "Set Member Limits"

Toggle the setting to "On" and you can set limits individually or set a standard amount for everyone by clicking the checkboxes for the corresponding students.

You can set the limit one by one for each student.

For convenience, you can apply the same monthly limit to all selected users using the Monthly Limit Amount setting.

Setting Up for a Smooth Class: Best Practices & Support

1. Pre-Prepare Workflows in Advance

To ensure a seamless experience for your students, set up your workflows at least a week before your class. Instructors are often busy with last-minute logistics, and the more workflows you have, the more time you'll need to configure them properly on ThinkDiffusion. Some workflows may be incompatible with others or become outdated, so early preparation helps avoid disruptions. We highly recommend a quick 20–30 min pre-test with our TD-Support team.

Pre-Preparation Checklist:

  • Install any custom nodes required for your workflows.
  • Install all necessary models.
  • If a custom node or model isn’t working as expected, contact TD Support immediately via your assigned support channel, i.e Discord.

2. Get Your Students Logged In Before the Class

To avoid delays during the session, ensure all your students have successfully logged into ThinkDiffusion ahead of time. This helps prevent any technical hiccups during the class.

Important Note:We do not automatically notify individuals you add to your account to prevent issues like mistyped email addresses causing spam or security risks.

  • Once you’ve added your team members, you’ll need to send them a personal email with one of the PDF instructions provided below to guide them through the login process. Once they sign up, they’ll automatically gain access to your workspace.
  • Additionally, please ensure your workspace has sufficient funds loaded to allow each student’s machine to launch seamlessly. Running out of credits mid-session can disrupt the learning experience.

PDF for school accounts running on Google email systems

PDF for school accounts running on NON-Google email systems

 3. Leverage TD Support

Stay in close communication with the TD Support team throughout your setup process and during the class. Your onboarding should include access to a dedicated Discord chat for quick assistance.

4. Notify TD for Scheduled Events

For scheduled training sessions or classes, inform the TD team at least one day in advance. This allows us to:

  • Ensure machine availability for your session.
  • Optimize file system performance for multi-user workflows.
  • Arrange high-performance compute machines and pre-warm them for instant launch.

5. On-Demand Priority Support

If you need real-time technical support during a session, coordinate with your TD advisor to have a dedicated support team available. This ensures any unexpected issues can be resolved immediately.

By following these steps, you'll create a smoother experience for both yourself and your students while maximizing the performance of ThinkDiffusion.

🚫 Don’ts (Follow These for a Smoother Experience)

When working on ThinkDiffusion machines, keep in mind that setting up the environment is slightly different from a local PC. To avoid issues, follow these guidelines:

ComfyUI Manager

All TD machines come with ComfyUI Manager pre-installed - there’s no need to reinstall it. Doing so can cause machine failures. Since ThinkDiffusion runs in a cloud environment and users don’t have CLI access, we preload the manager to ensure stability.

Adblocker Issues

Using an ad blocker can interfere with system requests, leading to frequent disconnects and errors. If you experience connection issues, try disabling your ad blocker.

Avoid Unnecessary Custom Node Installations

Each installed node adds to your launch time. If you have over 30 nodes, your machine may take 5+ minutes to start or even fail. Remove any unused or outdated custom nodes to keep things running smoothly.

Running Advanced Workflows on Lower-End Machines:

Some workflows (e.g., Flux, Hunyuan, LTX, Flux Lora, WAN, SkyReels, or other complex video workflows) require substantial GPU resources and may fail with Out-of-Memory (OOM) errors if you're not using a sufficiently high-end machine. For these workflows, use an Ultra Machine to ensure adequate VRAM and performance.

Uploading Large Files (Over 2GB)

If you’re on a Workspace or Team account, use the 24/7 file browser for files over 2GB. This lets you manage files without launching a machine, saving credits and reducing downtime.

Following these best practices will help you get the most out of ThinkDiffusion while avoiding unnecessary disruptions.

Support for Newer Workflows

ThinkDiffusion offers multiple versions of each app to ensure compatibility with different workflows, custom nodes, and models. These versions vary in stability and feature support, giving users flexibility based on their needs.

Available Versions on ThinkDiffusion:

  • Beta: The latest version, often based on the most recent commit from the official repo. Supports the newest custom nodes, workflows, and models but may have some instability.
  • Current: The stable release, typically two weeks to a few months old. Recommended for users who want a balance of stability and newer features.
  • Previous: A stable version, usually older than four months. Best for workflows that rely on older models or legacy custom nodes.

Each version comes with supporting notes to help you choose the right one for your workflow.

Common Issues

Machine Failed to Launch: 

There can be multiple reasons for launch failures.

Import failed on custom node

  • Check logs for the recently failed machine using the 24x7 file browser.

  • Navigate to comfyui/logs and open the latest failed machine log file.

  • Look for errors related to custom nodes or "Import Failed."
  • Remove the failing custom nodes from comfyui/custom_nodes and restart your session by launching a new machine.

Machine Took More Than 7 Minutes to Launch and Failed Without Errors in Logs:

  • If no errors are found in the logs but the machine failed on launch, the issue could be slow loading.
  • If you have more than 50 custom nodes, loading all nodes can be time-consuming.
  • In ThinkDiffusion, any launch exceeding 10 minutes is marked as failed.
    • Solution: Go to comfyui/custom_nodes and remove unnecessary custom nodes.

Machine failed for unknown reason

  • If neither of the above solutions work, you can reset your ComfyUI or A1111 environment from the Profile Menu.
  • Click on Profile > Locate Reset to Defaults > Click on Comfy/A1111 Reset.
  • This will clean only custom nodes/extensions while keeping your models and output files intact.

File Browser Not Working: There can be multiple reasons for file browser failures.

Machine is in an Out-of-Memory (OOM) State

  • Some workflows may consume excessive memory, causing the machine to crash and making the file browser unresponsive.
  • Solution: Upgrade to a higher-tier machine like ULTRA to run your workflow.

File Browser Still Not Working After Relaunch

  • This can be due to a dependency conflict caused by a custom node.
  • Example: Some older custom nodes may require outdated dependencies (e.g., typing package), leading to file browser failure.

If the issue persists, contact Support (Discord) for further assistance.

Custom Node Import Failed

Some custom nodes may have dependency conflicts or require additional Python packages.

Solution:

  1. Open Manager > Locate the failing custom node.
  2. Click Try Update.
  3. Restart ComfyUI.
  4. If the issue persists, contact Support for assistance.

ComfyUI Reconnecting/Disconnected

Possible Causes & Solutions:

  • Adblocker: Disable any ad blockers and refresh your browser.
  • Out of Memory (OOM) Exception: If ComfyUI is unresponsive, try using the queue feature or restart your session.
  • Installing a Custom Node or ComfyUI Restarting: Wait a few minutes before relaunching the session.

ComfyUI Grey/White Screen

Solutions:

  1. Refresh the browser or reload the page.
  2. Switch to a different network.
  3. If the issue persists, contact Support.

Uploaded Model Not Visible

  • If you uploaded a model but cannot see it under the custom node: 
    • Verify that the file size matches the original source and that the download has fully completed.
    • Refresh your browser. ComfyUI does not automatically update the list of models and custom nodes.

Not able to download gated/restricted models

  • Contact the Support Team to assist with downloading restricted models within your workspace.

Generation stuck or cancelled

Possible Causes & Solutions:

  1. Failure on Nodes or Missing Config/Files:
    • Check the node with a red border.
    • Ensure all required models are selected.
    • If the node requires input (image/video), provide the necessary file.
    • Verify that input/output connections are correctly linked.
  2. Out of VRAM/Memory:
    • Try launching an ULTRA machine.
    • If the issue persists, contact Support.
  3. Unknown Error:
    • Check logs under comfyui/logs for the root cause.
    • If unresolved, contact Support.

Node Not Found in Manager: Sometimes, the preloaded manager may not be fully synced with the upstream ComfyUI Manager Repository: 

Solution:

  • Open ComfyUI Manager > Click Install via Git URL.

  • Enter the GitHub URL of the custom node in the popup.

  • Click OK and wait for the installation to complete.

  • Once installed, click restart.

If the installation fails, reach out to Support on Discord for further assistance.