Studio, Transcribe, Notebook, Mindmap, Slides and Storyboard run on one sign-in and one credit balance. Write in Chinese and the reply comes back in Cantonese, with Hong Kong's weather, transit and markets built in.
Starter credits on sign-up. No card required.
One Google account, one credits pool — every app shares them. Start anywhere, switch without re-signing in.
The flagship workspace. Multimodal chat across 210+ models, live voice, multi-agent collaboration, deep research, image, video and music synthesis, a code workspace and persistent memory.
AI audio and video transcription from uploads or a YouTube URL, with speaker diarization, multi-provider routing, and export to text, SRT and VTT.
Source-grounded AI workspace. Upload documents, ask grounded questions, and generate briefs and notes with traceable citations.
Generate mindmaps from any topic and turn them into presentations. Edit nodes, restructure branches, and present straight from the browser.
Generate polished presentations from a single prompt. Pick a model, edit slide-by-slide, and export when ready.
Break a script into shots, generate first/last/reference stills, then render each clip.
Log every drink, track spending and sugar over time, and get AI recommendations based on what you actually order.
Track expenses, categorise spending automatically, and forecast next month with AI. Multi-currency and receipt capture.
Write, edit and play back music notation in the browser. Transcribe from a photo or audio, or compose from a prompt.
No update to install. Every app picks these up the next time you open it.
Live voice now has a photoreal avatar in 2D and 3D. It lip-syncs from a real photograph — a bundled portrait or your own.
Studio can now run the JavaScript it writes inside a locked sandbox and show you the result, plus new Canva and OneDrive connectors and a Calendar that sees every calendar you share.
On Auto, Studio picks how hard to think on each turn along with the model. Choose a level yourself and it is always respected.
MiniMax voices now read a whole reply over one continuous voice session, so sentences run into each other instead of stopping between them.
Long recordings are split and transcribed across several speech engines at once. File recordings into collections and search every word spoken.
Every citation opens the exact passage it came from. An audio source jumps straight to the second it was said in Transcribe.
Three new Studio tiles turn your sources into a slide deck, a narrated video overview or a one-page infographic.
Slides Polish now keeps a revision history with undo, exports Google charts, and generated slide images no longer drop out of a deck.
Shot prompts are written the way your chosen video model documents — Seedance, Wan or Veo — instead of one generic style.
A shared mindmap now arrives exactly as written, Chinese included. Links shared before the change still open.
Ask in Chinese and get spoken Cantonese back — 嘅, 咗, 喺, 嚟 — not textbook Mandarin. With Hong Kong's own data wired in.
Write in Chinese and Studio replies in spoken Cantonese, not Mandarin prose. No prompt engineering, no language switch, no reminding it every turn.
Live HKO weather and warnings, MTR and bus arrivals, Transport Department traffic, and HK plus US market quotes — read from the source, not guessed.
Gemini, GPT, Claude, Grok, Qwen, GLM, MiniMax, DeepSeek and Kimi on a single prompt surface. Top up once and spend it across every app.
No per-app setup. Sign in once and every module above is ready to use.
Use Google, Microsoft, GitHub, or a verified email or phone number. One account signs you into every app on this page.
Launch Studio, Transcribe, or any module above. Inside Studio, choose a model from the picker, or leave it on auto-router.
Every chat reply, transcription, or render draws from the same credit balance. Top up once in Billing, use it in any app.
Real-time weather, MTR, bus, and Gmail — Cantonese by default. Then slide decks, storyboards, research and code on top.
Responds in spoken Cantonese when you write in Chinese — no language-switch needed.
Planner, researcher, critic — orchestrated in parallel or sequence on a single task.
Multi-step retrieval across the live web. Synthesised answers with traceable citations.
Nano Banana Pro stills, Seedream 5.0, Gemini Omni cinematic motion, Seedance 2.5, Wan 3.0 and MiniMax H3 — up to 4K.
Live voice across Gemini Live, GPT Realtime, Grok Voice and Qwen Audio Realtime, 150+ named TTS voices.
Reason across millions of tokens. Context caching keeps your reference corpus warm between sessions.
Auto-captured facts and preferences, recalled across every chat. Inspect, edit, delete or pause it anytime.
Live HKO weather, TD traffic, MTR/bus schedules, and market quotes built in, plus OAuth bridges to Drive, Gmail, Notion, Outlook and GitHub.
Auto-generated quizzes and flashcards from any thread or file, with TTS narration in your chosen voice.
Auto-router picks the best model for the job — or pin a specific one. Routing is reversible per turn.
We treat your data the way we'd want our own treated. Three things you should know:
Full terms & policyYour sign-in identity — name, email and avatar from Google, Microsoft or GitHub, or the email or phone number you verify with a code — plus the content and usage records each app needs to work.
Chats and account data sit in our database on a server in Tokyo, behind Cloudflare. Files and generated media sit in Google Cloud Storage, encrypted at rest. All of it is scoped to your account and synced across your devices.
Never. No third-party brokerage, no ad-targeting, and no training of our own models on your content. The AI providers that process your requests follow their own terms.
Alchemys App (shown as SPCY on spcy.hk) is a Hong Kong–built workspace of AI apps that share one sign-in and one credits balance: Studio, Transcribe, Notebook, Storyboard, Mindmap, Slides and Reminders.
No. One account (Google, Microsoft, GitHub, or a verified email/phone) signs you into every app. Credits are shared across the whole workspace, so you can start in Studio and continue in Transcribe or Slides without signing in again.
More than 160 models across Gemini, GPT (including GPT 6 Astra), Claude, Grok, Qwen, GLM, MiniMax, DeepSeek, Kimi, Doubao Assistant and others. Per-model credit costs are listed on the public pricing page.
New accounts receive starter credits. Usage is billed in credits per model. See the pricing page for current per-model costs. Managing a balance requires sign-in.
Yes. The interface supports English, Cantonese (廣東話) and Simplified Chinese. Transcribe AI is built for Cantonese, English and Mandarin speech, including speaker diarization.
Our application servers run in Tokyo, behind Cloudflare. Chats and account data are held in our database; files and generated media are held in Google Cloud Storage, encrypted at rest; all of it is scoped to your account. Your prompts go to the AI provider that serves each request, which can be outside Hong Kong, including mainland China for some models and automatic modes. We do not sell your data or train our own models on it. See the Terms (Sections 6 and 17) for details.
Sign in with Google, Microsoft, GitHub, or a verified email or phone. Starter credits are on the house.