The Visuals Agent chat got calmer, and Visuals got a lot better on your phone.

Before, the Agent asked all its questions at once, wrote long walls of text, and you had to watch the visual canvas to see a run finish. Now it asks one question at a time, tells you in a line what it did, and shows the whole run right in the chat. On iPhone and iPad, Visuals fits the screen and responds to a single tap.

The short version:

Questions now show up one by one, in place of the message box, with a counter like 2 of 3. Close them and come back later without losing your answers, or press Skip questions. Typing your own message counts as an answer.

Some questions now say Select all that apply, for things like who a video is for, where several answers are true. The Agent's recommended answer is always first. Skip a question and it uses its recommendation.

While the Agent works you see one plain line about what it is doing. After each turn you see how long it thought and a short summary like "Added 3 cards, 2 connections". Tap it to see each step. Scripts and briefs show as a small document card with an Open button.

The message box is simpler too: attach, song and settings in one field. While the Agent is busy, the send button queues your message, and queued messages sit on top with Send now. On a computer you can keep the chat as a side panel, full screen or floating.

When a run starts, from the chat or the visual canvas, the chat shows a live run card. It lists every card in the run with its model and main settings. Each one gets a spinner, then a check or a failure mark, and its result fills in as it lands. The finished video sits at the top. Open any finished result in the full viewer, and edit text results right there.

Above the message box, a small chip keeps track of the flow:

The Run all button now shows its price, the same way the test video does. Credit numbers for video are estimates until the run finishes.

A lyric video could show one font and line breaks in the editor, and different ones in the export. This mostly happened with fonts that only live on your own computer, like Futura, Arial or Helvetica.

Now the editor and the export use the same font files, sizes and spacing, so lines break the same way. Fonts that only work on some computers are no longer in the picker. Older projects that used them switch to a close free match, for example Futura to Jost and Arial to Arimo, and the picker shows which one a project now uses.

The picker now has 260 fonts, all free to use in your videos. New ones include Manrope, IBM Plex Sans, Barlow, Urbanist, Bodoni Moda, DM Serif Display, Geist Mono, Unbounded, Syne, Bungee, Monoton, League Gothic, Rock Salt, Kaushan Script, Sacramento, Michroma, Oxanium, Teko, Cinzel and Metal Mania. They work in the caption and text style panels, in preview and in export.

On the Image and Poster pages, the Generate button could show a lower price than the real one, for example 8 credits for an edit with a reference image that really cost 19. Then the request was turned away.

Now Visuals checks the real price when you press Generate. If it is higher, or the quality has to change, a Review the price and settings window shows the images, model, quality and credits before anything starts. Choose Generate for the shown credits, or Keep editing. No prices changed.

New video cards now start on Seedance 2.0 Fast: 16:9, 8 seconds, sound on, 720p. Cards you already saved keep their model, and the Visuals Agent still picks the best model for each shot.

The Agent also has 10 new playbooks: Lip-Sync, Character & Voice Continuity, Ad Versions, Event Posters & Announce Kit, Thumbnails & Video Covers, Identity Sheet, Product Photoshoot, AI Creator Persona, Motion & Character Transfer, and Mini Dramas & Drama Ads. Turn them on or off under Skills in the Agent's settings.

Find the connection details in Settings, API & MCP.