AI Assistant

Vū Studio includes a built-in AI assistant that helps you find content, generate new media, and automate tasks — all from a conversational interface inside the app.

Opening the AI Assistant

Click the AI icon in the left sidebar, or press the AI button in the library toolbar. A chat panel opens on the right side of the screen.

You can also start a chat from the Home Screen: type your question in the search bar and choose the Ask AI row at the top of the suggestions, and a new Vū AI chat opens with your question. See The Redesigned Home Screen.

What the AI Assistant Can Do

Search Your Library

Ask natural language questions to find content:

• "Show me backgrounds with ocean themes"
• "Find the Samsung launch video from March"
• "What images do I have with warm sunset colors?"

The assistant uses visual embeddings (AI-generated representations of each image's appearance) to find visually similar content — not just filename matches.

Generate Images

Request new AI-generated images directly from the chat:

• "Generate a futuristic cityscape at night for my tech conference background"
• "Create a subtle abstract texture in deep blue and gold"

Generated images are saved to your library automatically.

Edit Images (Inpainting)

Upload an existing image and ask the assistant to modify specific parts:

• "Remove the logo from the bottom right corner"
• "Replace the sky with a sunset"

Use the paint brush tool in the attachment panel to mark the area you want changed.

Generate Video

Request short video clips via RunwayML integration:

• "Create a slow-motion pan across a forest floor"
• "Generate a looping abstract particle animation in red and black"

Answer Questions

Ask about Vū Studio features, get help with workflows, or request step-by-step instructions for anything in the platform.

Take Actions in the App

Vū AI can now perform a wider range of in-app actions on your behalf and is more aware of what you're currently working on, feeding relevant in-app information back to you as you work. Ask it to carry out supported tasks directly from the chat instead of switching between screens.

Suggest Your Next Move

Vū AI can also propose something without being asked. A small glowing orb sits in the bottom-right corner of the app and lights up when it has a move worth offering — usually something that would otherwise take several clicks and drags.

• Where it appears — anywhere in Vū Studio except the Vū AI page itself, where the conversation is already the channel. Drag the orb to another edge (left, right, or bottom) if it is in your way; it stays where you put it.
• What it does — an idle orb is a quick way into Vū AI: click it to open the search palette in AI mode and type a request. When a suggestion arrives the orb brightens and its panel opens with the proposal and three choices — Accept, Not now, and Don't suggest again. Accepting runs the work for you and streams the progress inside the panel.
• Quiet — the toggle in the panel header stops suggestions from opening on their own. The orb still glows when it has one, and you open it when you are ready.
• The first suggestion — so the feature is easy to find, one arrives within the first minute of using the app: if you have not connected any outside apps yet, the orb offers Connect apps, which is what gives Vū AI the context behind later suggestions.
• Displays — when a suggestion involves a display or a wall, it names the specific display you are working with rather than referring to a display in general.

Expanded In-App Actions

The AI assistant can perform a broad range of actions inside Vū Studio on your behalf — creating and managing scenes, files, and assets, and automating most of the in-app tasks you would otherwise do by hand. If there's something you'd normally do manually while building or organizing content, you can ask the assistant to do it for you.

Supported action areas include:

Area	Examples of what you can ask
Scenes	Create a new scene, update an existing scene, organize scenes into folders
Files and assets	Upload, rename, move, tag, or delete media; generate new assets and add them to your library
Content generation	Generate images and video, edit or expand images, and save the results directly to your library
Organization	Group content, tidy folders, and carry out routine library housekeeping

The goal is that most in-app tasks you would automate by hand can be requested in chat instead. Ask the assistant what it can do in your current context if you're unsure whether a specific task is supported.

Project Chats: Linking a Chat to a Folder

A chat can be tied to one library folder, which turns it into a project chat. Vū AI then treats that folder as the project the conversation is about: it searches the folder before the rest of your library, files everything it creates straight into it, and — because access comes from the folder — everyone who can edit the folder can use the chat.

Starting a project chat

There are two ways in.

From the folder. Open the folder in your library at one.vu.studio/library — click the Vū Studio icon to go Home, then click Library — select the folder, and choose Chat with Vū AI from the actions. The first time you do this the folder's project chat is created; after that the same action reopens it. You need owner or editor access on the folder.

From a new chat. In a chat that has no messages yet, open the + menu in the composer and choose Link to a folder. Pick the folder and select Link folder.

A project chat opens on an empty thread that tells you what is different about it: whatever you make here gets saved here, and it looks in this folder first. That panel disappears as soon as you send your first message.

What linking does, and does not, allow

caution

Linking is permanent and one-time. A chat can only ever be linked to one folder, and the link cannot be changed or removed afterwards. Linking also makes the chat readable by everyone who can edit that folder — which is why you can only link a chat that has no messages in it yet. If you could link a chat that already had history, you would be publishing that history to a new audience after the fact.

Once a chat is linked, a bar across the top of it reads Project chat · files into [folder name], with two controls on the right:

• People — opens the folder's own sharing settings. Sharing the folder is how you add someone to the chat, so there is one place to manage both.
• Open folder — jumps to the folder in your library.

Who can do what follows the folder:

	Folder owner / editor	Folder viewer
Open and read the chat	Yes	No
Send messages	Yes	No
Pin assets in the attachments panel	Yes	No
Rename or delete the chat	Creator only	No

Folder viewers get no access at all — not even read-only. A viewer who could send messages would be using Vū AI to write into a folder they can only look at.

If you did not create the chat, a line under the bar explains why you are in it and how to leave: your access comes from being able to edit the folder, so leaving means removing yourself from that folder's sharing.

Deleting the folder does not delete the chat — it reverts to being a private chat belonging to whoever created it.

Working in a shared chat

One person at a time. While Vū AI is working on someone's request, nobody else can send. If you try, you are told a reply is already in flight rather than getting an error — wait for it to finish and send then. This is deliberate: two requests landing at once would have the second undo the first, and every reply runs with one person's account and connected apps, so there is no sensible way to answer two people in one breath. Holding the turn is also what lets everyone else read the exchange and ask something informed.

You can see who said what. In a project chat each person's messages carry their name and profile picture. Your own still say You.

You can watch a reply arrive. While someone else holds the turn, the chat shows what Vū AI is doing — for example vu.ai · replying to Judah… — and each completed step of a multi-step answer appears as it finishes. Only the person who sent the request sees the answer typing out word by word; everyone else gets it in stages.

Each turn runs as the person who sent it. Connected apps and tokens are the sender's, so a teammate's request uses their own connected accounts, not yours.

Personal memory is switched off. Vū AI will not save anything to your personal memory from inside a project chat, and it says so plainly if you ask it to. Its memory would otherwise change the assistant's behaviour depending on who happened to be typing, and something you told it privately could surface in a chat your whole team reads.

What Vū AI knows about the folder

Three things change once a chat is linked:

1. It knows what is in the folder. The folder's name and a sample of its contents are part of the conversation's context, so you can refer to things in it by name.
2. It searches the folder first. Ask for something by name and Vū AI looks in the linked folder and its subfolders before the rest of your library. If it finds nothing there it widens the search rather than telling you your library is empty.
3. It files its output there. Images, microsites, canvas designs and animations created in a project chat go into the linked folder automatically — including when you are browsing somewhere else at the time. The link is a deliberate choice about what the chat is for, so it wins over wherever you happen to be.

Automations

An automation is a task you describe once and Vū AI runs for you afterwards — either on a schedule ("every weekday at 5pm, list the action items from today's meetings and email them to me") or when something happens in a connected app ("when I get an email from my manager, Slack me").

Automations live at one.vu.studio/ai/automations. Click the Vū Studio icon to go Home, click Vū AI, then click Automations in the Vū AI sidebar.

Automations are not playback scheduling

This page is for tasks you ask Vū AI to carry out. To schedule content onto a display — an asset, folder, scene, or a Sleep/Wake command — use the schedule instead (see Scheduling Playback). The Automations page carries a link across to it.

Creating an Automation

1. Open one.vu.studio/ai/automations (Vū Studio icon → Vū AI → Automations).
2. Describe the task in the prompt box, including when it should run — a time and a repeat for a schedule, or the event that should set it off.
3. Submit it. Vū AI works out whether it needs a schedule or an event trigger and sets it up.

If you are not sure what to automate, the page offers starter cards, including one that asks Vū AI to suggest automations based on the apps you have connected and how you work.

Vū AI settles the details before it saves anything. Any app the task needs must be connected first (see Connected Apps), and anything the task will act on has to be pinned down to a specific item while you are still in the conversation. So if you ask it to "update a microsite with the numbers each Monday," expect it to ask which microsite — or whether to make a new one each time — rather than working it out on the night. Answering that question once is what stops the automation acting on the wrong thing later.

Previewing an Automation Before It Saves

Before a scheduled automation is saved, Vū AI offers to run it once on today's real data and show you what it would have delivered — for an email automation, the email itself — without sending anything. Taking the preview is your choice: say no and the automation is saved anyway. Vū AI always asks, though, and it offers a new preview if you later change what an automation does.

A preview changes nothing:

• It only reads. It cannot cast to displays, move assets, schedule playback, or create anything.
• Nothing is sent. The message the automation would deliver is shown to you instead.
• Connected-app steps are skipped. Steps that use your connected apps — reading Gmail or posting to Slack, for example — are left out, and the preview says what it would have done there.
• It is free. Previews are not charged against your token balance.

Rewording an automation after its preview does not force another one, but changing where it delivers (switching email to Slack, say) or what it acts on does. A preview is good for two hours. Automations set off by an event in a connected app ("when I get an email from…") are not previewed, because the event they wait for has not happened yet.

Time Zones

Automations run in your local time, so "today" and "this afternoon" in an automation mean your day, not UTC.

• When you set up a schedule, Vū AI compares your device's time zone with your primary calendar's. If they disagree — because you are travelling, or your computer's clock is set wrong — nothing is saved until you tell it which one to use.
• When it confirms the schedule, it names the zone, for example every weekday at 5:00 PM Eastern Time (America/New_York), so a wrong zone is easy to spot.
• A recurring automation keeps the zone its schedule was set up in. One-off and event-triggered automations use your primary calendar's zone when they run; with no calendar connected, they fall back to UTC.

Meeting Action Items and Recaps

When an automation asks for action items, recaps, or notes from your meetings, Vū AI reads the Vū Intelligence sessions that recorded those meetings — your calendar entries carry no meeting content of their own. Each meeting on your calendar is paired with the session that captured it, and meetings that were not recorded, were missed, or have not happened yet are reported as such rather than as meetings with no action items. Action items mean work to be done after the meeting, listed separately from decisions and open questions.

Managing Your Automations

Your automations are listed under Your automations on the same page, each with its schedule or trigger and its recent run history. Every entry offers:

Control	What it does
Run now	Runs the automation immediately, without waiting for its next occurrence. A run usually takes between about 30 and 90 seconds.
Edit prompt	Opens the instruction for editing. Change it and select Save, or Cancel to leave it alone.
Show prompt	Expands the entry to show the full instruction as it will run.
Cancel	Cancels the automation so it stops running.

Editing a recurring automation changes every future run — so this is how you say "point that at the other microsite" or "move it to 8am" without rebuilding the automation from scratch. An automation that happens to be running at that moment cannot be edited; wait for the run to finish and edit it then.

Knowing Whether an Automation Worked

Each entry shows the status of its recent runs, and Vū Studio emails you as well.

• When a run finishes cleanly, you get a summary email — "<name> finished running" — with what the run produced. This one is optional: clear Email me a summary when an automation finishes under Automation Options on the Automations page to turn it off, or use the unsubscribe link in the email itself.
• When a run could not do what it was asked, you get a separate email saying so, naming what failed. This one always sends and cannot be switched off, because an automation that quietly stops working is the failure worth knowing about — the usual cause is something the automation depended on being renamed, moved, or deleted since you set it up, which only you can put right.
• When a run works but finds nothing, you still get its delivery, not a list of blanks. A short result is always delivered as asked; only an entirely empty one is replaced by a note of what the automation checked, why there was nothing, and what would change that. This counts as a successful run, so no failure email is sent.

A failed step does not always mean a failed run: an automation that does four things and misses one reports the miss and still delivers the rest. Runs that fail partway are still charged against your token balance, because the work up to that point was really done — see AI Tokens.

Context Awareness

The assistant is aware of the tools, pages, and apps available inside Vū Studio and of where you are as you work. It can surface relevant in-app information back to you dynamically — including links to the right tool or page for what you're trying to do — so the guidance you get fits your current context.

Clarifying Questions

When a request is ambiguous or the assistant needs more detail to proceed, it can ask you a clarifying question and present options to choose from, rather than requiring you to type out every detail. This speeds up creative workflows where you're working with tools and images and would otherwise have to enter a lot of text.

When an Asset Reference Cannot Be Resolved

The same applies when Vū AI cannot work out which asset you meant — for example a video request that names an image the assistant cannot match to anything in your library. Rather than quietly generating from an unrelated asset, it stops and asks which asset you meant. Point it at the right one (attach it, or name it precisely), or repeat the request without a reference asset so it generates from your prompt alone.

Voice Mode

Voice mode lets you talk to Vū AI out loud and hear it answer, inside the conversation you are already having. It is available in any Vū AI chat — the dedicated AI page at one.vu.studio/ai (click the Vū Studio icon to go Home, then click Vū AI), and the chat sidebar of a Vū Intelligence session.

Turning Voice Mode On

1. Open a Vū AI conversation.
2. In the message composer, click the microphone button beside the text box — its tooltip reads Talk to vu.ai (voice mode).
3. A voice panel opens above the composer. The microphone button stays highlighted while voice mode is on; click it again (Turn off voice mode), or the × in the panel, to close it.

Turning voice mode on does not start a different kind of conversation — it is the same chat, with the same history, attachments, and tools.

Speaking a Request

The panel offers two ways to talk, chosen with the switch on its right:

Mode	How it works
Hands-free	The microphone stays on and your request sends automatically after you pause for about a second. Speaking while Vū AI is talking interrupts it and starts your turn.
Tap to talk	Tap the microphone to start, speak, then tap again to send.

The panel shows the current phase — Tap the mic to start, Listening…, Thinking…, Speaking… — alongside a live caption of what is being heard, so you can watch the words being recognized before they send.

Whatever you say is transcribed and posted into the same conversation as an ordinary text message. It appears in the transcript as text, and it is answered exactly as a typed message would be: a spoken turn is a normal turn whose input happened to come from a microphone.

Hearing the Answer

Replies are spoken as they stream in, sentence by sentence, so you hear the answer begin instead of waiting for the whole reply to finish. Spoken answers are also shaped for the ear — short, one idea at a time, without markdown, lists, or read-aloud URLs — while the full written reply still appears in the conversation.

• The speaker button in the panel toggles spoken output (Speak replies aloud / Mute spoken replies). Voice input still works with spoken replies muted.
• While Vū AI is talking, a Stop button next to the phase label cuts the audio immediately, without you having to say anything.

Browser Support and the Text Fallback

Voice mode uses your browser's own speech recognition, which works best in Chrome and Edge. Where the browser does not support it, the panel says Voice mode needs a browser with speech recognition (try Chrome) and nothing else changes — the conversation carries on as text, and you can close the panel and keep typing.

On iOS, hands-free listening is not available: the mode switch is hidden and the panel tells you to tap the microphone to talk, then tap again to send.

AI Tokens

AI features consume tokens from your plan's monthly allocation. Generation tasks (images, video) use more tokens than search or questions. Monitor your usage on the Account page (profile menu → Account).

Team plans have a shared token pool — all team members draw from the same balance.

Buying More Tokens

If you run low, you can top up without changing plans. On the Account page, find the AI Tokens card and click Buy More, then complete checkout. Purchased tokens are added to your balance — for team profiles, they're added to the shared team pool.

Attachments

You can attach images or files to your chat messages for the assistant to work with:

• Attach a screenshot for the assistant to help recreate or modify it
• Attach a video for analysis or suggestions
• Attach a reference image as a style guide for generation

Asset Actions

Every asset in the conversation — and every tile in the Attachments panel on the right — carries a small action bar across its top. It stays hidden until you hover the asset, so it never covers the artwork; on a touch screen the actions are always visible, so a tap reaches them directly. From it you can:

Action	What it does
Pin to top	Keeps the asset at the top of the Attachments panel.
Reference this Asset	Attaches it to your next message.
Play this Asset	Casts it to a connected display.
View in Library	Opens the asset where it lives in your library.
Open fullscreen mode	Expands the asset over the page.

Resizing the Chat and Attachments Panels

On a desktop-width screen both side panels can be resized. Move the pointer to the inner edge of the conversations panel on the left, or of the Attachments panel on the right, until the cursor changes to a resize handle (its tooltip reads Drag to resize), then drag. Each panel keeps the width you give it.

Tips

• Be specific about the look you want — instead of "make a background", try "make a dark, moody industrial warehouse with dramatic side lighting"
• Use the library search conversation style for finding existing assets rather than using the search bar — the AI understands context better
• For inpainting, be precise when drawing the mask area — smaller, well-defined selections give better results

Uploading Files to Chat

Getting files into Vū AI is faster and more flexible. You can drag and drop files straight into a message, attach several files at once, and hand the assistant a document — a PDF, a spreadsheet export, or a plain-text file — so that it can work from what is inside it rather than from your description of it.

What you can add

Dropped onto the message box, or chosen with Upload a file in the + menu, Vū AI accepts:

Type	Extensions	How it is handled
Images	the image formats your library accepts	Uploaded to your library, then referenced in the message
Video	the video formats your library accepts	Uploaded to your library, then referenced in the message
PDF	.pdf	Uploaded to your library; the assistant reads the text out of it on request — see PDFs
Text and CSV	.txt, .csv	Attached to the conversation only — not added to your library — see Text and CSV files

Anything else is turned away with "Only images, videos, PDFs, and text files (.txt, .csv) can be added here." PowerPoint is the one exception worth knowing: a .pptx cannot be dropped onto the message box, but the assistant can read one that is already in your library — see PowerPoint decks.

You can also paste an image straight into the message box. Nothing is uploaded until you send, so a pasted screenshot you decide against costs you nothing.

Drag and drop a file

1. Open a Vū AI conversation — either the AI search overlay in the top navigation, or the dedicated Vū AI page at one.vu.studio/ai (click the Vū Studio icon to go Home, then click Vū AI).
2. Drag one or more files from your computer over the message input. A drop-zone highlight appears to show the message will accept the files.
3. Release to drop the files onto the message. You can also use the + button to the left of the message box if you prefer to browse — see The Plus Menu.
4. Add any text prompt describing what you want done, then send.

What happens on send

For images, video, and PDFs, Vū AI uploads the file to your library and then immediately references the uploaded asset in that same message — there is no separate upload step. The assistant can then work with the real library asset directly for edits, style references, generation inputs, and other supported tasks.

• Multiple files can be attached to a single message.
• Uploads honor your plan's token and upload limits, the same as uploading through the library.
• Let one upload finish before starting the next; while an upload is in flight, a further attachment is refused with "Hold on — finish the current upload first."

Text and CSV files are the exception: nothing is uploaded, and they never appear in your library.

Text and CSV files

A .txt or .csv file attaches directly to the conversation. Vū AI reads it in your browser and sends the contents along with your message, so there is no upload wait and no new asset cluttering your library. Ask about it the way you would ask about anything else — "what are the action items in these notes", "which rows in this export are missing a venue", "turn this list into a microsite".

The chip clears when you send. The attachment shows in the message box while you are composing, then moves onto that message in the transcript, where it stays. It does not reappear in the message box when you come back to the conversation later — you do not need to re-attach the file to keep asking about it.

How much of the file the assistant keeps in view depends on its size. Up to about 6,000 characters (roughly 1,000 words) the whole file goes into the conversation, so every follow-up sees all of it. Above that, Vū AI writes a short overview into the conversation instead — for a CSV, its columns, its row count and a few sample rows; for prose, what the document is and how it is laid out — and keeps the full contents to hand so it can go back and look up specifics when you ask for them.

Limit or caveat	Detail
Largest file	500,000 characters. Anything larger is refused with "…is too large to attach (over 500k characters)."
Empty file	Refused with "…is empty."
On your other devices	The overview travels with the conversation, so you see it wherever you sign in. Looking a specific value back out of a large file works only in the browser you attached it in.
Sums and counts on a CSV	Lookups and "what does this data say" questions are dependable. Exact totals and counts across a large sheet are not — the assistant is reading the sheet, not calculating over it. Use a spreadsheet for the arithmetic.
Deleting the conversation	Clears the stored contents along with it.

PDFs

Once a PDF is in the conversation, ask for what is inside it: "summarize the brief I just dropped in", "pull the section headings out of that whitepaper", "build a microsite from this report". You can also ask about a PDF that is already in your library, by name.

• While it reads, the assistant shows Reading the PDF…, then reports back what it read — the file name and its page count.
• Scanned and photographed PDFs work too. A PDF with no text layer cannot be read as text, so Vū AI falls back to looking at the pages, and tells you it is doing so: No text layer found — reading the PDF visually…. This takes longer than a normal read. PDFs above 15 MB are not read this way; for those, a text-based PDF is the reliable route.
• A very long PDF is read up to roughly 32,000 characters, and the assistant tells you when the contents were truncated at that point.
• If nothing can be pulled out at all, it says so rather than guessing — "Couldn't extract any text from … — it may be blank or unreadable."

Reading a PDF's contents and placing a PDF are different requests. If you only want the document to appear on a page or in a microsite, ask for it by name — the assistant does not need to read it to put it somewhere.

PowerPoint decks

The assistant can read the text out of a .pptx presentation — slide titles, body copy, the structure of the deck — which makes a deck a practical starting point for a microsite or a summary.

A .pptx cannot be dropped onto the message box. Upload the deck to your library first (see PowerPoint Presentations), then bring it into the conversation with Attach an asset in the + menu, or simply name it in your request.

The text comes back laid out slide by slide — Slide 1:, Slide 2:, and so on — so you can ask about a particular slide. Slides that hold only images or shapes and no text are absent from what the assistant reads; if a deck's content lives in its pictures, describe it in your prompt or attach the images themselves.

The Plus Menu

The + button at the left of the message box opens a short menu with three ways to bring something into the conversation:

Option	What it does
Upload a file	Opens your computer's file browser; the file is uploaded to your library and referenced in the message.
Attach an asset	Opens the asset picker so you can attach something already in your library.
Connect apps	Opens the connections screen, where you connect the outside apps Vū AI can act on for you.

Choosing a Model

Vū AI lets you choose which underlying model it uses for a given task, so you can match the model to the job (for example, a different model for image generation than for video). By default, each task runs the model you've selected — you don't have to specify one every time.

Where to set your model preferences

Open vu.ai Settings and select the Models tab. (Settings also includes a Memory tab.) The Models tab lists each AI task and lets you pick the model that task should use. Your choices are saved to your account and persist across sessions.

Tasks you can choose a model for

Each task has its own chooser and a sensible default:

Task	Default model
Generate image	Flux 1.1 Pro
Edit image	Nano Banana Pro
Expand image (reframe)	Luma Reframe
Upscale image	Topaz
Animate image (image to video)	Google Veo-3

Additional tasks, including remove background and text to video, also have their own model choices.

Available models

The models offered depend on the task. Examples include:

• Image generation / editing: Flux 1.1 Pro, Imagen 4, Nano Banana Pro, Nano Banana 2, Seedream 4, GPT Image 2, GPT Image 1.5
• Video: Google Veo-3, Runway Gen-4, Pruna P-Video

Choosing a model for a single request

You don't have to change your saved default to try a different model once. You can name a model directly in your chat request (for example, "generate this image using Imagen 4"), and the assistant will use it for that request only. You can also ask the assistant to change your saved default for a task, which updates the same preference as the Models tab.