Hi, how can we help you today?
Hi, how can we help you today?
As you train custom models in Scenario, effective management becomes essential for maintaining an organized workspace, especially as your library grows to dozens or even hundreds of models. Model management involves organizing and documenting your custom models to ensure they remain easily accessible and quickly usable - helping you track versions, share with team members, and quickly find exactly what you need for any project. This section covers the best practices to effectively manage your models on Scenario
Name each model with a descriptive, specific title that indicates its purpose, style, or subject. Add the version number if needed. Titles are the first elements you retrieve with a search (Cmd/Ctrl+F)
In addition to the name, add comprehensive descriptions in the “Details“ tab of your model page. You can include important information such as:
The model's purpose or intended use
Effective prompt structures that work well with this model (in addition to the ones in “pinned images)
Any special techniques used during training
Examples of successful outputs or use cases, including links if applicable
Clear descriptions are especially valuable in multi-user environments where others may use your models.
Having visual references for your models will significantly improve navigation and retrieval of models. Set a representative thumbnail for each model by selecting your best generation that showcases what the model does best (you can also upload any image from your computer).
Add multiple pinned images to highlight different capabilities of your model. To add pinned images, hover over an image you like, and click the “pin“ icon to the top-right. Or you can find the “pin” icon to the top bar on each image page:
Pinned images serve three important purposes:
They create a carousel in the thumbnail view, giving a quick preview of various outputs, for you and others in the workspace
They provide reference for what your model can typically produce
They enable and improve the accuracy of the “Prompt Spark” feature, which can suggest new prompts based on pinned images (See dedicated article)
Custom tags are efficient tools for filtering and finding models quickly. You can add relevant tags in the “Overview” and “Details“ tabs of the model page. For multi-word tags, use hyphens (e.g., "character-design").
Consider tagging models by:
Style (cartoon, realistic, 3D, pixel art)
Subject (character, landscape, product, UI)
Project name (game_project, marketing_campaign)
Other technical aspects
Save time and tag multiple models at once by selecting them in the model gallery, before applying tags to the whole batch (use the “tag” icon to the bottom-left).
More information about Tags on our dedicated article.
Collections help group related models together. You can create collections based on projects, styles, or subjects, and add models to them from the model page (via the top bar) or by selecting multiple models in the gallery (via the bottom-right menu).
Important: when a model is added to a collection, all outputs generated with this model will be automatically included in the same collection. This might be convenient to automatically group related outputs into one place, saving time without the need for manual sorting.
More information about Collections on our dedicated article.
Once you've established effective prompt structures for a model, it’s possible to add frequently used style tokens or descriptors as “prompt embeddings“ in the Details tab. These embeddings will be automatically included in every prompt used with the model.
Prompt embeddings ensure consistency and save time by eliminating the need to repeatedly type the same tokens.
You can also automatically analyzes your training images and creates a style description, in the prompt embedding section. Just click on Describe btton in the prompt embedding box. Especially valuable for difficult-to-train styles or for balancing prompts that might pull toward different aesthetics.
You can efficiently share and manage models across workspaces or with team members:
Transfer Models: Move models between workspaces by navigating to the model page, clicking the three-dot menu, and choosing "Transfer Model." Enter the target workspace ID or pick it from the list.
Export Models: To export a model, select "Export" from the model's options (3-dot) menu. This creates a downloadable file that serves as a backup or can be shared externally. Once initiated, you'll receive a notification via email with a download link. Additionally, a small red dot will appear in the "Recent Tasks" section, located in the top-right corner of the interface. Click the link in the list to download your exported model directly.
Import (Upload) Models: You can easily import custom models directly to Scenario. Access the upload feature through two paths: either via the left-hand menu under Model > Upload (or Model > New > Upload).
You have two primary methods for importing models:
Direct Upload: Upload a .safetensor file directly from your device. Scenario supports various model types, including SDXL LoRA, SD1.5 LoRA, Flux Dev LoRA, and SDXL/SD1.5 Checkpoints.
External Import: Paste a model from Hugging Face by providing the model name (e.g., stabilityai/stable-diffusion-xl-base-1.0) or import from CivitAI by pasting the model URL.
After confirming the import, the uploaded model will be processed in the background. You'll receive email and on-screen notifications once the upload is complete. Uploaded models will be available in your Model Gallery.
As your model library expands to hundreds of custom AI models, streamline your workflows with these best practices:
Use consistent naming conventions across your workspace, especially in team settings.
Document successful multi-model workflows to simplify replication for yourself and others.
For large teams, designate someone to manage model organization and set standard practices.
Track related models: When retraining, keep both versions briefly to compare outputs, then delete the weaker one after deciding.
For multi-LoRA models, tweak compositions by opening the model page, selecting "Edit" from the three-dot menu, and adjusting weights or swapping components.
By implementing these good practices, you'll create a well-organized, easily navigable model library that enhances productivity and collaboration, even as your collection grows to hundreds or thousands of custom models.
Update Model Details: Scenario API - Update Model
Upload Model: Scenario API - Upload Model
Manage Model Tags: Scenario API - Manage Tags
Duplicate a Model: Scenario API - Duplicate Model
Transfer Model: Scenario API - Transfer Model
Organize Models into Collections: Scenario API - Collections
Was this helpful?
Quentin