A photoreal face for Live voice
Live voice can show a photoreal avatar in 2D or 3D, animated from a real photograph. Use a bundled portrait, or upload your own photo and it rigs itself the same way.
Ask a question, upload a file, generate an image or video, or talk to it out loud. Studio routes what you type to the right model and the right tool, so you never have to know which one to pick.
Sign in to start — every new account gets a free credit balance shared across all Alchemys apps.
What's newUpdated October 2026
Six things Studio does now that it could not a month ago.
Live voice can show a photoreal avatar in 2D or 3D, animated from a real photograph. Use a bundled portrait, or upload your own photo and it rigs itself the same way.
Ask for a calculation or a quick script and Studio runs the JavaScript it wrote inside an isolated sandbox — no network, no files, capped memory and time — then shows you the output.
Auto read-aloud with MiniMax voices now streams the whole reply over one continuous voice session, so sentences flow into each other instead of pausing between them.
Find, create, export and autofill Canva designs, read files from OneDrive, and let Calendar see shared and secondary calendars — not just your primary one.
On Auto, Studio now picks how hard to think on each turn together with the model. Set a level yourself and it is always respected.
Tell Studio — typed or in Live voice — what you bought, and it prepares an editable card for Drink Record or SmartSpend. Nothing is saved until you confirm, and never twice.
No commands to memorise — say what you want, and Studio picks the tool.
Ask questions, get code reviewed, or have something written, then switch between Gemini, GPT, Claude, Grok, DeepSeek, Qwen, MiniMax, GLM and more without starting over.
Describe what you want, get it made, then ask for changes — pictures, short clips with sound, and full songs.
Deep research reads dozens of sources, cross-checks them, and writes a report that links back to every one.
Open live voice for a real back-and-forth, have it interpret between languages as you speak, or jam on music in real time.
Split one job across several helpers working in parallel, or send the same question to a few models and pick the best answer.
Reach into Google Drive, Gmail, Notion, GitHub, Slack and more once you connect them, and chat with Studio from Telegram.
About five minutes to your first useful answer.
Sign in
Open Studio and sign in — new accounts start with a free credit balance shared across every Alchemys app.
Start a new chat
Click “New chat” in the sidebar. The title gets written for you from your first message.
Pick a model, or don't
The model picker sits at the bottom-right of the composer. Choose “Auto” at the top of its list and the best model for your question is picked for you — the button then reads “Auto-select”.
Type, talk, or attach a file
Type your message, then send it with Ctrl+Enter (⌘+Enter on a Mac) on a computer, or Enter on a phone — you can change this under Send Shortcut in the Settings panel. Tap the microphone to dictate instead, or use the “+” button to attach a file or a YouTube link.
Read, switch, and carry on
The answer streams in as it's written. Under it you get Copy, Regenerate and More options (Read aloud, Rewrite, Fork from here), and you can Edit your own message. Not happy with it? Switch models mid-chat — everything so far carries over — or use Model Arena to send the same question to a few models side by side and continue with whichever answered best.
Every part of Studio, and what lives there.
The picker next to the composer lists the main models. The full list, grouped by provider and by what each one is for, is under Select Model in Settings.
Reads your question and picks the model for you — the safe default.
A planner splits the job into up to 12 steps, helpers work through them four at a time, and everything is merged into one answer. Find it under “Others” in the picker.
Sends the same question to two or three models at once so you can compare answers and continue with the one you prefer. Find it under “Others”.
Always free — hops between free models until one replies. Good for quick, everyday questions.
An open model for everyday questions, without the premium tier. Find it under “Others”.
The picker next to the composer shows the main families — Gemini, Dola, Doubao Assistant, GPT, Claude, Kimi, Grok, MiniMax, DeepSeek, MiMo and Hunyuan. Open Select Model in Settings for the full list, including Qwen, GLM, Mistral, Perplexity and more: filter by provider or type (text, image, video, music, TTS, agent), search by name, and sort by capability or price.
Small models are fast and cheap for everyday questions; larger ones earn their keep on reasoning, long writing and code. Some models — and all media generation — spend credits. See pricing, and top up on billing.
Studio isn't limited to what it already knows.
Use the “+” button next to the composer. Local File adds a document, image, video or audio clip from your device (you can also drag a file into the window). Uploaded Files reuses something you already uploaded, and YouTube Link adds a video by its address. On a phone the menu also offers Take Photo, Record Video and Album. To keep a file's content in memory, open User Memory and choose Attach file.
Studio can look things up on the web. Use the Search skill for a question, or switch on Google Search, Google Maps and URL Context in the Settings panel — those three work with Gemini models only. Other models search the web and read links with built-in tools. Research Mode in the settings menu makes that search deeper but slower.
Teach it once instead of repeating yourself every session.
Studio can save notes about your background and preferences and bring the relevant ones back in later chats. Switch this on or off with “Allow model to remember”. Read, add, edit or delete memories in User Memory (settings menu) or on the Personalization page. “Access chat history” lets the model see earlier turns and search your other chats when a question needs them.
A skill is a saved way of working that you attach to one message. Pick one from the Skills button next to the composer, type “/”, or use the gallery on the Personalization page (Mine, Built-in, Templates), where you can also create your own or have AI generate one. Upload documents to your Knowledge Base and Studio can search them when a question fits — switch that on with “Auto-search knowledge base”.
Connect outside tool servers — the Model Context Protocol standard — under MCP Servers on the Connections page, so Studio can reach your own systems.
Every answer can be heard, and you can just talk back.
Every reply has Read aloud under its More options menu. In the Settings panel, switch on TTS Voice Output and choose a read-aloud model — Fish Audio, Gemini, MiniMax, OpenAI, Qwen or xAI voices — or turn on Auto-read responses in the settings menu so new replies play themselves. You can also create your own voice on the Personalization page for Fish Audio and MiniMax. Tap the microphone next to the composer to dictate instead of typing.
Open Live Chat from the sidebar for a real-time voice chat you can interrupt. Turn on Camera or Share Screen and it sees what you see, and choose a 2D or 3D Face. Switch Mode to Translation to speak and hear your words translated aloud, or to Music to steer a real-time music session with weighted prompts. Save the session as a chat when you finish. Live uses credits while it runs, at a per-minute rate, and the session closes if your credits run out.
Set it up once and let it run without you.
On the Personalization page, open Reminders & Routines → Routines and add a routine: a prompt that runs on a schedule. Choose where results go — in-app as a chat, Telegram once you've linked it, or a webhook event (routine.completed). After three failures in a row a routine switches itself off.
Under Reminders & Routines → Reminders you can add reminders, sort them into folders and show completed ones. Set Quiet hours (From and Until) so reminders stay silent, or open the Reminders app for the full view.
On the Connections page, open Outbound Webhooks, add your URL and choose an event: Chat created, Chat message, Expense created, Expense updated, Transcription completed, Score saved, Slides generated or Notebook created. Add a secret if you want the calls to be signed, and use Test to send a sample.
Deep research, video and music generation run in the background, tied to your account rather than the device you started them on — start something on your laptop and pick it up on your phone later, with a notice when it's done.
Keep related work together, and reach into the apps you already use.
Group related chats into a folder and start new chats straight inside one; share a whole folder as a read-only link. Choose Attach Notebook in the “+” menu to cite a Notebook's sources in a chat and pull its outputs back in as attachments.
Open a chat's menu to Pin, Rename, Move to folder, Duplicate, Regenerate Title, Share or Delete it. On any reply, More options → Fork from here starts a new chat from that point and leaves the original alone. Share link makes a read-only link that expires after seven days and can be revoked sooner; you can also copy the chat as text or export it as Markdown, JSON or PDF. Temporary Chat in the sidebar saves nothing at all.
Connect a service on the Connections page and Studio can use it in chat. Type “@” in the composer to attach a record from another Alchemys app, or to point at a connected service. To chat from your phone, link Telegram on the same page with a link code — Telegram chats show up in your history but stay read-only in Studio, so you continue them in Telegram.
The specifics, in one place.
Worth knowing
Don't paste passwords, card numbers or other secrets. Studio gets things wrong sometimes — verify anything that matters, and treat its answers as a starting point, not professional advice. Generated images carry an invisible provenance watermark, and pages simply don't exist for anyone who isn't signed in.
Sign in, ask it anything, and see which model answers best.