Changelog

A handful of smaller wins. Add people straight from an editor, share a single manuscript chapter, and let people know when they have been given access to a project, with a switch to turn those notices off. A new collaborator lands in the editor they were invited to. The screenplay and manuscript editors open over the conversation, the screenplay revision menu docks beside your selection, and read aloud uses the audio player and only shows its controls once the voice starts.

A reliability pass across the chat and the production pipeline. You can type while a reply is still running, and up to five messages queue up behind it. A retried turn no longer repeats your question, the tools, or the transcript. The branch you switch to stays put, and an edited opening no longer carries the abandoned branch into the reply. Attached pictures reach the story engine, survive a refresh, and no longer get cut off.

Productions hold together better. The screenplay writes its scenes in steady batches and reports each part of the outline as it starts and finishes. Character, location and prop lists fill in from the screenplay. A cut refuses to leave a shot out, a run that comes up short says so and keeps its place, and a music video now starts a short film beside the music. One conversation can start a second production of a different kind.

The Usage page is rebuilt from the ground up. It shows what you have used, where it went by project, and how it has moved over time. Usage now runs on two windows, a five-hour limit and a weekly limit, and unused resets bank for later. The studio only mentions your allowance when it is actually running low.

Every generation is metered at what it really cost, so the numbers on the page match the work that was done. Buying extra credits and resets from the page is on the way.

Every conversation now carries a row of agents above the chat. Each agent is a small mark colored by the kind of work it does, and it glows until it finishes. Hover one to see what it is doing in plain words. Open the row into a full view of every agent at work, sized to the conversation, on laptops and tablets as well as desktops. Finished agents step out of the way.

Runs are sturdier underneath. One agent that hangs can be stopped without stopping the rest, and a cancelled agent restarts on its own without restarting the whole run.