Release notes

While anything is generating, a capsule appears next to Share: «2 running · ~3 min». Open it for each task's progress and remaining time, and click one to jump straight to that node — across canvases too, so you can keep working elsewhere while it renders.

See 1.2.

Each turn now ends with a Branch action: it copies the conversation up to that turn into a new chat and switches over — the original stays untouched, so a wrong turn no longer needs to be wrestled back.

See 12.1.

The composer Translate button is no longer English-only: open its menu and pick English / Chinese / Traditional Chinese / Japanese / Hindi; click again to bring your original back.

See 10.4.

The "Camera move" chip on the video composer opens a full library: official moves play their preview on hover, a click applies one; keep favorites, and save your own wording under "My moves".

See 4.2.

The audio node toolbar gains two buttons: Smart split cuts the clip into new cards by phrase and beat; Custom split takes a sentence like "keep 0-2s" or "delete 10-15s" and cuts locally — both free.

See 5.6.

«Save to my assets» on the canvas now opens a dialog: one row per item — choose the folder, give it a name you'll recognize, add tags, then save. No more silent bookkeeping.

See 10.1.

The previz desk (formerly the director desk) gains an animation timeline: lay motion tracks for characters, have the camera follow someone, or write one sentence to stage it all — Space to preview, then record frame by frame into an mp4 from the top right, saved as a file or landed on the canvas as a video node.

See 8.8.

Creating a Skill no longer means filling a form: the Agent asks a few questions right in the chat (start from scratch, from the current canvas, or update / self-test an existing one), then writes and saves it itself. A .md file or a whole folder you already have can be imported too.

See 12.2.

The image composer gains Presets: fifteen one-click presets that set up the wording and aspect for standard pieces like character sheets, scene sheets and three-view turnarounds — you say what you want, it writes the long part.

See 3.6.

Image and video nodes get a full spread of models at once: photo-grade FLUX.2 Pro, design-oriented Recraft, audio-synced Veo, the arena-topping Happy Horse — pricing is aligned per vendor, so the composer's estimate is truer than ever.

See 14.8.

The video node model list becomes a full wall: Kling O3 (the only family that takes video in, video out), Wan 3.0 and Happy Horse are here — and every parameter is aligned with upstream, so the options you pick are no longer silently dropped.

See 14.8.

No more guessing before a batch script: dotv model lists every available model (filter by node type or keyword), and dotv model --key <model> spells out each parameter's type, allowed values and default.

See 14.7.

The Language submenu now offers five: Simplified Chinese, Traditional Chinese, English, Japanese and Hindi — interface, docs, legal pages and the desktop app all switch together.

See 14.2.

The assistant on the canvas that does what you say is now uniformly called Agent — buttons, panels and docs all say the same name; the work it does is unchanged.

See 12.1.