Business introductions
Collect an accurate description, choose a short scene and create a visual introduction. If you start with a website URL, build an extraction and review step first: a video model does not automatically read the website.
Build an AI video product
Give video generation a useful workflow.
Let a customer choose a direction before spending credits on a clip. Build the brief, review and result library around their business, instead of exposing a wall of model settings.
Demos start with prepared briefs and shot suggestions. You choose when to generate; preparing a plan does not spend AI credits.
Make a product, not just a prompt box.
A business introduction tool can start with a reviewed business brief, offer a few scene directions, and generate the chosen clip. A course app can create a visual for a lesson. A campaign tool can keep alternate takes beside the same product. Bubble supplies the application experience; fal supplies the selected model’s generation.
Product campaigns
Keep the product reference, chosen scene and output format together. Use an image-to-video endpoint when the real product must appear; a text-only prompt cannot guarantee its identity.
A library of reusable clips
Store approved clips by project, campaign or lesson. Let customers select an existing result before they pay to generate another take.
A practical brief-to-clip sequence
Approve the scene
Describe one subject, one main action and one camera movement. Keep claims, spelling and business details in a reviewed script rather than asking the model to invent them.
Submit once and track the job
Send the chosen endpoint its required inputs and save the job reference. Show a waiting state and a status action while the video is processing.
Review the actual clip
Check subject consistency, motion, artifacts and the requested duration. Save an accepted file. Treat voiceover, exact captions and final assembly as separate workflow steps.
Video model inputs are not interchangeable
A text-to-video model may accept a prompt and aspect ratio. An image-to-video model also needs a source image; other endpoints may require first and last frames or a source video. Duration, resolution and audio options can change the price. The managed custom-model path uses the endpoint’s own input structure, so a model selector needs to update the form as well as the model name.
Choose a model that fits the job.
Search the fal catalog and connect compatible queue inference endpoints through a model ID and model-specific JSON inputs. Running a request requires schema validation, provider access and verified pricing. A catalog listing is not a guarantee that the model can run: models without a supported usage quantity stay unavailable until configured. The service does not promise that every catalog model has been tested.
Realtime sessions, streaming generation and training are separate integrations. A new model may need different input fields and result handling; choosing its name alone does not adapt your interface.
Technical references: fal queue and result documentation · Bubble server-side plugin actions.
One managed plan
Build it. Run it. Get support.
$49.99 / month
$30 in AI credits included.
Plan details & account →Subscription availability is shown on the account page.
- Free Bubble connector for the managed service
- Building skill for adapting an existing Bubble app
- Support from The Wise Bubbler
- Tracked generation jobs and a shared app credit allowance
Credits reset monthly with no rollover. Your app’s users share its allowance. Eligible requests use verified fal unit prices and supported usage quantities. Credits are reserved before generation and reconciled to actual provider costs afterward, which can take over an hour. Different models and settings use different amounts. There are no automatic overage charges.
Bubble hosting, your AI assistant and any other services your app uses are separate. Review the selected model’s terms for your intended use.
Before you build
Does one generation produce a finished advertisement?
A generated clip is an asset. A finished advertisement can also need approved copy, voiceover, music, captions, a logo and editing. Build and test those additional steps for your product rather than promising them from a single clip request.
Can I generate vertical video?
Use an endpoint that supports the aspect ratio or resolution you need. Check its input schema before offering a portrait option. Cropping a landscape clip later may cut off important subjects.
Why use a queued job?
Video can take longer than an ordinary page action. Submitting a queued request and keeping its job reference lets your app check for the result later without submitting the same creative request again.