File Management

Most AI tools make you re-upload files or burn compute just to grab an output file or check a log. That gets expensive fast - and it wastes your time.

🗂️ Folder Structure

When you launch a machine on ThinkDiffusion, the platform automatically creates a top-level folder for that specific app. Each app has its own folder. You’ll see these in the file browser.

File paths matter. Some workflows will fail if a model isn’t in the right subfolder.

Example top-level folders:

  • /comfy/ → for ComfyUI machines

  • /kohya/ → for LoRA training

  • /automatic1111/ → for A1111 sessions

Each of these app folders contains its own structure - and they’re separated so you don’t mix tools.

Inside each app folder, there are common subfolders:

/logs/

Where machine logs are stored. Each time you launch a machine, it writes logs here. File names match the machine ID in your URL.

/models/

Where model files are stored (e.g. LoRAs, checkpoints). Depending on the app, these might be split into:

  • /models/checkpoints/

  • /models/loras/

  • /models/vae/

/outputs/ or /outputs/img2img/

Saved images, video frames, or render results show up here - depending on the app.

/input/ or /images/

Where uploaded images, masks, or reference files go. Some workflows look for inputs in specific folders.

Note: On free accounts, files are only accessible while a machine is running. Files are automatically deleted after 48 hours of inactivity. The 24/7 File Browser is available on TD Pro accounts.

🗂️ 24/7 File Browser

  • Your files persist between sessions

  • You don’t need to launch a machine just to grab a LoRA

  • You can prep and organize workflows without paying for idle GPU time

  • You can debug logs even while your machine is launching

With the 24/7 File Browser:

The 24/7 File Browser gives you full access to your files - without launching a machine. You can upload models, check logs, organize folders, and download outputs, all without paying for GPU time. It’s only available with ThinkDiffusion Pro.

Last updated

Was this helpful?