Skip to content
Cask Manual

Generation in Cask Film

Generation produces an asset. It does not produce an edit.

A finished generation lands on the canvas as a card, and in the library. It has no position in time, so export cannot see it. Drag it onto a timeline track to put it in the video. This is the same rule as the canvas and the timeline, and generated shots are where it bites hardest.

Picking a model

Section titled “Picking a model”

Different models take different inputs. An image-to-video model expects an image bound to a specific slot: a start frame, an end frame, or a reference attachment. The app only offers the slots the model declares, so a model that takes no start frame does not show one.

Sending an image to the wrong slot is rejected before anything is charged.

Chaining shots

Section titled “Chaining shots”

A sequence holds continuity when each shot starts on the previous shot’s last frame. Two routes get you there:

  • Continue the finished shot directly, which extracts its last frame and uses it as the next shot’s start frame.
  • Extract the frame yourself, then start a generation with that frame in the start-frame slot.

Either way it is a fresh generation and costs like one.

Regenerating

Section titled “Regenerating”

Regenerating a card runs the job again at the same price. The previous result is not kept for free.

When a generation fails

Section titled “When a generation fails”

Read the provider’s own reason rather than the summary line. The summary often suggests adjusting your prompt for failures that have nothing to do with the prompt: an unsupported aspect ratio, a rejected input slot, a provider outage.

A failed generation is refunded, and the card records whether it was: None used, Credits refunded, or the number actually spent.

Section titled “Related”