AI Chat and Voice Mode

Chat is the primary way you interact with AI on Scholarly. You can ask questions, get explanations, generate study materials, and have full voice conversations with AI Chat.

Chat Basics

Click Chat in the sidebar to start a new conversation. Type your question and the AI responds in real time.

• Conversations auto-save and appear in the sidebar for easy access later.
• You can queue messages while the AI is still responding. They combine into a single prompt and send automatically when the current response finishes.

Leaving a Chat Mid-Answer

You don't have to sit and watch a response finish. Close the tab or move to another page and the answer keeps running. When you reopen the conversation, it reconnects to that same answer and picks up where it is — it does not start over.

Stop response still works the way you expect: it ends the active response, whether you started it in this tab or reconnected to it.

Model Selection

The named model catalog includes GPT 5.6, Gemini, Grok, Deepseek, Mistral, and Claude options. AI Chat is unlimited on Premium and Laureate, on every model in that catalog.

Free includes 3 AI Chat messages total. This is a lifetime allowance: it follows your account and does not reset daily. A message is charged only when an AI response is saved; a failed response returns the reserved message. Retrying a response creates another AI response and uses another message.

Premium's recommended named choices are GPT 5.6 Terra, Grok 4.6, Claude Sonnet 5, and Gemini 3.1 Pro Preview. Paid plans can also choose Gemini 3.8 Flash, Mistral Medium 3.5, and Mistral Large 3.

GPT 5.6 Sol and Claude Opus 5 are creation models, not Chat models. You can pick them when you generate research, slides, video lectures, podcasts, and infographics, but they do not appear in the Chat model picker.

Auto uses GPT 5.6 Luna on every plan: low reasoning on Free, high reasoning on Premium and Laureate. See Choosing an AI Model for the full plan-by-feature defaults table.

The model picker keeps things simple: Auto is the default, a short Recommended Models list covers strong named choices, and the rest live under More Models. Each model has different strengths. On Free, paid models appear with a lock badge so you can see what an upgrade unlocks. You can also switch models mid-conversation and the AI carries over the full context. Scholarly remembers the model you pick for AI Chat separately from your creation tools, so starting a new conversation picks up your last choice instead of resetting to Auto.

Thinking

Enable the Thinking toggle in the model menu to have the AI reason through problems before replying. This produces better answers for complex questions, math, and multi-step logic.

When thinking is on, you see a live progress summary while the model reasons. Once finished, the summary collapses into a compact "Thought for X seconds" block.

Some models always reason and do not show the toggle -- thinking is built in. Gemini 3.1 Pro Preview and Grok 4.6 are examples.

Auto manages reasoning for you. On Premium and Laureate it runs GPT 5.6 Luna at high reasoning, so you get careful answers without touching the toggle.

Grounded Answers

When you ask about something you've linked or uploaded, the assistant sticks to your actual material. It won't invent menus, fabricate quotes, or add citations it can't back up — if it isn't in your source or in a real web result, it says so. If chat ever feels off on a tough question, switch to a different model (see Retrying with a Different Model below) or open the source directly to compare.

Linking Your Content with @

Type @ in the chat input to link your existing Scholarly content directly into the conversation. The AI reads whatever you link and uses it to give contextual, grounded answers.

You can link:

• PDFs
• Pages
• Flashcard sets
• Podcasts
• AI Meeting Notes
• Research sessions
• Folders

Whatever you attach — a page, a folder, a recording, a file — appears in the conversation as a chip with its name on it the moment you send, so you can see exactly what the AI was given.

Chat Knows Your Library

You don't have to remember the exact name of a file to bring it into chat. The assistant can search your library by name, type, or date, and pull up the matching item -- ask things like:

• "What was that PDF I uploaded last week?"
• "Find my biology lecture recording."
• "Open my last research report on neural networks."

You can also tell the chat to read one of your items. It loads the PDF, recording, podcast, or video (and uses the transcript for audio and video so it can answer faster) and uses the content to answer your question -- handy when you want a summary, an explanation of a tough section, or to compare two of your files. Linking with @ still works for pinning exact files into the conversation; library search is the faster path when you're not sure of the name.

Google Drive in Chat

You can pull files straight from Google Drive without downloading and re-uploading them. Open the content menu in the chat box and click Google Drive for a one-click picker. Choose the exact files you want and they drop right into the message box, ready to send.

Once you connect Drive, the AI can use your Google Docs, Sheets, Slides, PDFs, Word docs, PowerPoints, and text files as grounded context for its answers. See Connections for how to connect or reconnect your account.

File Uploads

Upload files directly into chat by attaching images, PDFs, documents, and other files. Each uploaded file appears as a pill in the conversation. Click any file pill to preview it in the side panel.

Text and data files you attach in chat (notes, CSVs, code, calendars, and more) are saved to your library as editable library files and linked as sources — so they stick around after the conversation ends. Open one and you can preview it, edit it inline, or reuse it as a source for any AI tool.

File Preview Panel

Files linked or attached in a chat (notes, code, CSVs, text, images, calendars, and more) now open in the preview panel right next to the conversation — the same way PDFs do — instead of popping out in a new tab. You can keep reading the file and the conversation side by side without leaving the chat.

The preview panel is resizable (drag the left edge) and supports many file types:

• PDFs -- Full document viewer.
• Images -- Inline display.
• CSVs -- Clean spreadsheet table with row numbers and sticky headers.
• Code files -- Syntax-highlighted with line numbers (30+ languages).
• Markdown -- Rendered with full formatting.
• HTML -- Live rendered preview.
• SVG -- Visual rendering with animation support.
• Audio files -- Inline player (MP3, WAV, OGG, etc.).
• Video files -- Inline player (MP4, WebM, MOV, etc.).
• Calendar files -- Event preview grouped by date with time, location, and description.

Navigate between all files in a conversation with prev/next arrows or the file picker dropdown.

Fullscreen Files Sidebar

In fullscreen chat mode, open the Files button in the header to see all files the AI has generated or referenced in the conversation. Browse, preview, and download files without scrolling through the chat history.

Save Generated Files to Your Library

When the AI creates a file in chat — a code sample, a CSV, a Markdown outline, a diagram — most outputs include a Save to library action. One click turns it into a real library file you can return to anytime: it opens in its own page, syncs across Home, search, the sidebar, folders, and Pinned, and can be reused as a source for flashcards, podcasts, video lectures, worksheets, study guides, and more.

AI Tools

The AI has access to several tools it can use during a conversation:

• Web search -- Search the internet for current information and citations. When you ask about something time-sensitive — today's weather, current prices, live scores — chat fetches the latest results from the web instead of relying on cached pages.
• Code execution -- Run Python code for calculations, data analysis, and working through problems step by step.
• Image generation -- Generate diagrams, illustrations, and cover art on demand. See the Image Generation guide for details.
• Image editing -- Edit images you upload or that the AI generated. Combine multiple reference images in a single editing request to blend or restyle multiple sources.
• Document creation -- Create PowerPoint presentations, Word documents, and PDFs directly from chat. Ask for slides on any topic, generate essays or reports as Word documents, or create formatted PDFs for assignments.
• Flashcard creation -- Generate a flashcard set from the conversation content.
• Podcast creation -- Generate a multi-speaker podcast directly from chat. Say "make a podcast about X" and it handles the rest. Completed podcasts play inline in the conversation.
• Video lecture creation -- Generate a narrated video lecture from chat. The video plays directly in the conversation when ready.
• AI Slides creation -- Generate a slide deck from chat. Completed slides link to their full view.
• Deep Research -- Start a full research session from chat. Say "research X" or "do a deep dive on Y" and the AI asks what to produce (a cited report, slides, a brief, or other files), then runs research in the background. The completed work appears inline when ready.
• Canvas -- Open a freehand drawing canvas to sketch diagrams, equations, or concept maps. Clicking Done saves the sketch as an image attachment the AI can analyze. See the Canvas guide for details.
• PDF page reading -- Read specific pages from any PDF you have linked with @.

When the AI runs multiple tools in sequence, they are grouped into a single collapsible block instead of showing each one individually.

When the assistant draws a diagram or flowchart in its reply, it appears as an openable file. Click it to view the full diagram, or download it to use elsewhere.

Retrying with a Different Model

If a response misses, hover over the AI message and click Retry. Pick a different model and the new variant streams in below the original. Both versions stay attached to the conversation, and a small switcher (for example, 1 / 2) lets you flip between them. Whichever variant is selected becomes the context for follow-up messages, so you can branch the conversation by switching variants.

Two limits worth knowing:

• You cannot retry while the original response is still streaming. Wait for it to finish, then click Retry.
• Each message holds up to 4 variants. Once you have four, Retry is disabled for that message — follow up in the chat or edit your question instead.

Slash Commands

Type / in the chat input to open a command menu with keyboard navigation. Slash commands let you trigger specific AI actions quickly.

Built-in commands include /new (start a new conversation) and /create-skill (create a custom skill). Predefined study skills like /summary and /study-guide are also available.

You can create your own custom slash skills with personalized prompts. See the Slash Commands and Custom Skills guide for full details.

Text Selection Menu

Select any text in a chat message and a floating menu appears with three actions:

• Ask -- Adds the selected text to your chat input for a follow-up question.
• Copy -- Copies the selected text to your clipboard.
• Speak -- Reads the selected text aloud.

Pinned Conversations

Pin important conversations to keep them easily accessible:

• Click the pin button next to the chat title.
• Hover over a chat card on the /chat page and click the pin icon.
• Right-click any conversation to pin it.

Pinned chats show as highlighted rows under the Chats tab in the sidebar, alongside your recent conversations -- the same row style used on Home. Pin state syncs across sessions.

Sharing Chats

Share any AI conversation with a public link. Click the Share button in the chat header to toggle visibility, invite by email, or copy a direct link. Shared chats display the full conversation read-only.

Background Tasks

When you ask the AI to create flashcards, video lectures, podcasts, or research from chat, these tasks run in the background. Track progress on your Home page where all active and completed tasks are visible. Click any task to jump straight to it -- even tasks that are still processing are clickable so you can watch progress in real time.

Changed your mind? Tell the chat to stop and it cancels the podcast or video lecture it is still working on. Say "stop that" or name the one you mean, and it confirms what was cancelled.

Product Feedback

If you tell the chat about a bug, a rough edge, or something you wish Scholarly did, it asks first before passing that along. Only when you say yes does it send the note to the Scholarly team, and a small Feedback noted card shows you exactly what was shared.

When You Hit a Lifetime AI Creation Limit

If you use your 1 free lifetime AI creation credit while building something from chat, the conversation shows a one-tap upgrade card right where the creation was queued — no dead-end error message, and no need to leave the conversation. Free AI creation credits do not reset. Cancel out and the chat continues normally.

Dictating a Message

Don't want to type? Click the microphone icon next to Send in the chat input. This is different from Voice Mode below -- it does not start a conversation with the AI, it just turns your speech into text so you can send it yourself.

• Click the microphone icon and the input turns into a recording bar with a live waveform, so you can see you're being heard.
• Talk, then click the stop button when you're done.
• Your words appear as text in the input box -- review it, edit it, or add more before sending. Nothing is sent automatically.
• Changed your mind mid-recording? Click the X to cancel and discard it.

Voice Mode

Click the microphone button to start a voice conversation with the AI.

Choose from Scholarly's searchable voice library. Luna is the default, and you can switch voices mid-session from the voice bar or set your preferred assistant voice in Settings.

That preference also controls read-aloud audio for chat replies, flashcards, and selected text. Voices for generated podcasts and video lectures are selected separately when you create them.

Control playback speed from the voice bar — choose 1x, 1.25x, 1.5x, or 2x. The AI voice stays natural at all speeds with pitch preserved.

Voice and text share the same conversation. You can switch between typing and talking at any point without losing context. Typed messages during voice mode are sent through the voice system -- the AI responds with audio instead of a regular text reply.

While in voice mode, the AI can still search your PDFs, notes, and the web to answer your questions.

Free users get 15 lifetime minutes of voice chat, and that allowance does not reset. Premium users get 2 hours per day.

For a detailed guide, see Voice Mode.

Recent Chats

Every content page shows a strip of your most recent conversations at the top of the chat panel. Click any entry to switch to that conversation instantly. If you have more than 5 previous chats, click Show more to see the full list.

This makes it easy to maintain multiple threads on the same content -- for example, one chat for chapter summaries and another for practice questions.

Your chat list on the Chat page shows which AI model answered in each conversation, next to that provider's icon, and Chat results in quick search show the same label. It always reflects the most recent active answer, so switching models mid-conversation or selecting a different retry updates it.

Chat Breadcrumb

Conversation headers include a Chat breadcrumb that takes you back to the main chat page (a clean starting point with no conversation pre-loaded). Refreshing inside a conversation keeps you on the same chat URL so you do not lose your place.

Context-Specific Chat

Chat is available on every content type in Scholarly, not just from the sidebar. When you open chat from a PDF, flashcard set, meeting notes, podcast, or research session, the AI automatically receives context from whatever you are viewing.

You do not have to link anything to start. On a video lecture, PDF, flashcard deck, podcast, recording, research session, folder, file, page, project, or the image playground, an empty chat already has that source in front of it -- the panel tells you so, and you can ask your first question straight away.

The chat panel starts open on new accounts. Show or hide it whenever you like and Scholarly remembers your choice for next time.

On PDF pages, the AI knows which specific page you are on and can reference surrounding content without you needing to link anything manually.

On a flashcard deck, chat can do more than answer questions — ask it to add, rewrite, or remove cards and it edits the deck for you. The changes appear in your study session immediately and are saved back to the deck. See Flashcards, Quizzes, and Exams for more.

One-Click Suggestions

When you chat with a source, one-click suggestions appear so you can turn the conversation into study material without typing a prompt — build Flashcards, a Practice Quiz, or a Study Guide straight from what you've been discussing, plus podcasts, video lectures, slides, infographics, mind maps, and data tables. Tap a suggestion and the creation starts as a background task.