Models and workflows
Browse the AI Models window, search and filter the catalog, compare models, read a model card and switch back to a model you used.
FlashBoards sends your shots to image, video and audio models. You pick one in the AI Models window, then set its inputs and run it. For the full flow, see Generate.
Open AI Models
Click Models in the bottom toolbar. It shows None until you pick one. The window has three tabs.
- Latest: recently added models and workflows, grouped by date. Click one to select it.
- Models: single models. Fast, with a fixed price per run for most.
- Workflows: multi-step recipes. More control and often cheaper, but they can take longer and the price depends on the run.
Click a card to make it the active model, then close the window. Variants such as Fast or Omni are separate cards with their own behavior and price.
Search and filter
- Type in Search models... (or Search workflows...) to match names and descriptions.
- Use the type filters for text, image, video and audio.
- In the left panel, pick a Base Models entry or a category. Hide filters panel gives the cards more room. On phones, the panel starts closed; use Show filters panel to open it.
- Click the star (Mark as favorite) on a card, then Favorites to show only those.
- Click the eye (Hide model) to hide a card. Show hidden models brings them back; Show only archived models (with the count) switches the list to retired models only, which you can still run. Turn it off to see the rest again.
- When discounts are running, the % button (Show discounted models) filters to them.
- Ratings in the left panel compares models by Arena score and Flash Vibes. Click a row to select that model.
Read a model card
A card shows the model name, a badge such as Official, Cheap, Stable, Fast, Legacy or Archived, what it does (for example Image to Video), the estimated time, a success rate and the price. Hover the price for a breakdown.
The success rate tells you how reliably the model has run lately, not how good it looks. Cards marked Under Review are greyed out and can't be selected.
Prices and discounts change, so trust the card and the Generate button over old screenshots. See Billing.
Click the i on a card for Model Details, or the external-link icon for the model's page on the Explore site.
Workflows
Click View workflow details (the i) on a workflow card to see its versions, the Workflow Inputs it needs and a GPU Tier choice. Plus runs on a faster machine than Default and costs more. The Form tab lets you set up and preview a run. Some workflow cards carry an NSFW tag.
Some workflows use a fixed prompt: the prompt box shows Prompt locked for this workflow or No prompt required, and you only fill the labelled slots such as Target Image and Source Face.
Audio models
Audio models are in the catalog too: music, speech, sound effects, voice changing and voice cloning, stem separation, voice isolation and audio description. Use the Audio type filter or search to find them. They take an audio item or text as input, and the result lands on the board as an audio item. Audio descriptions and new voice IDs land as text notes. Voice-cloning models need Voice consent ticked, and the person whose voice it is must agree.
Audio settings use plain labels such as Duration (seconds) and Instrumental only. ElevenLabs music and sound effects are priced by the second of audio you ask for, so a short clip costs less than a long one.
Return to a model you used
Right-click empty board space and hover Favorites or Recent to switch model without opening the window. Star models in Favorites (the star in the bottom toolbar) to keep them at hand. To restore a whole earlier setup, see Reusable generations.
Note
Changing model clears the board's input roles. See Inputs and references.
Choose a model
| You have | You want | Look for |
|---|---|---|
| Text only | An image | A text-to-image model |
| One image | A video | An image-to-video model |
| Several images, video or audio | One consistent shot | A multi-reference model |
| A character image and a voice track | A talking character | A talking-avatar model |
| A finished clip | Higher resolution | A video upscaler |
| A finished clip | Smoother motion | Frame interpolation |
For quick tests, pick cards marked Cheap or with a low estimate. For final quality, compare the cards in the category.
To browse by job, use Categories in the left panel of AI Models. Examples are Cinematic Video Generation, Concept Art & Style Exploration, Character, Fashion & Face Tools and Upscale & Restoration, which holds upscalers such as Topaz Starlight. Each category shows how many models it has, and only categories with models are listed. All Models (or All Workflows) shows everything again.
Match inputs to slots
Each model lists the slots it needs. Images fill image, reference and frame slots, videos fill video slots, audio fills audio slots, and text notes fill text slots. See Inputs and references. Models have stricter file limits than the board; see Supported formats.