Background Generations

AI generation takes time — anywhere from 30 seconds for fast models to 15+ minutes for heavy models. modelBridge handles this by running every generation in the background, freeing you to keep editing while results are being produced.

How Background Handoff Works

When a generation takes longer than expected, the plugin automatically hands it off to background processing. You do not need to do anything — the Generate button resets and becomes available for new work immediately.

Background polling continues for up to 30 minutes. If fal.ai accepted your request, modelBridge will catch the result — even if you switch models, browse the catalog, or close and reopen the panel.

The Active Generations Panel

A persistent panel sits at the bottom of the plugin, tracking all active background generations regardless of which model card is currently open. It survives model switches — you never lose track of a running job.

What Each Row Shows

Each generation row displays:

Status indicator — orange (running), gray (queued), green (completed), red (failed)
Model name — which model is generating
Elapsed timer — real-time counter since submission
Current step — what stage the generation is at
Expand chevron — click to see detailed progress

Expanded Progress View

Expand a row to see a 5-step progress tracker:

Sent — request submitted to fal.ai
Queued — waiting for GPU allocation (shows queue position when available: “Queued #2”)
Generating — the AI model is producing output
Downloading — retrieving the result file
Importing — adding to your Premiere Pro project

Completed steps turn green with a checkmark. The active step pulses orange. Step labels update in real time.

The expanded view also shows your input details — prompt text (truncated to 80 characters), duration, resolution, aspect ratio, and the model endpoint — so you always know which generation is which.

Notifications

When a background generation completes or fails, a floating snackbar notification appears with quick-action buttons. You can act immediately or dismiss the notification and come back later — completed results stay in the panel.

Completed Generations

When you navigate to a model whose background generation just completed, the row auto-removes from the Active Generations Panel and the model card shows the completed result. No manual cleanup needed.

Failed Generations

Failed rows stay visible with a red status indicator and a See error link. They are never auto-dismissed — you always see what happened. Each failed row can be manually dismissed when you are done reviewing the error.

Error messages in the panel use the same color-coded categories as error banners:

Red — input quality issue (fix your media or parameters)
Amber — billing or content policy (action required)
Blue — temporary infrastructure issue (usually resolves on retry)

Generation Time Estimation

After three successful generations with any model, modelBridge starts showing estimated generation time before you click Generate — for example, “~45 sec” or “~2–3 min.” These estimates are built from your own usage history, not hardcoded data.

During generation, the progress messages adapt:

“Almost done…” when you are close to the estimated completion
“Still working…” when the generation takes longer than usual

The estimates are median-based and rounded up slightly, so the actual time usually comes in under the estimate.

Recovery

Background generations are resilient:

Panel close/reopen — active generations continue and are picked up when the panel reopens
Model switching — the Active Generations Panel tracks all jobs regardless of which model is currently selected
Long generations — polling continues for up to 30 minutes with adaptive intervals (faster at first, slower over time)

If a generation exceeds the 30-minute polling window, a notification explains that the result may still be processing on fal.ai but the plugin has stopped checking. You can find the result in your fal.ai dashboard.