GIGIlistening…
Downloads

Get GIGI on your devices

Use the same account on desktop, browser, and phone. Add the hashtag.space Chrome extension for #protocol browsing, install the desktop app, or add the mobile site to your home screen with the QR code.

Mobile web & PWA

QR code — open on your phone

Scan with your camera app. Opens the map in the browser; use your browser's Add to Home Screen (Chrome / Safari) for an app-like fullscreen experience. The site registers as an installable web app where supported.

  • Same maps, portals, and sign-in as the website—no App Store build required for this flow.
Preview the mobile layout on this computer →

Scan with your phone camera

https://hashtag.org/portals/map

hashtag.space · Chrome

Browser extension

The official hashtag.space plugin brings #protocol navigation and discovery into Chrome—aligned with the same spatial web story as Geo Portal and Web3 bridge content on this site.

  • Resolve and open #domains from the toolbar; works alongside your normal browsing.
  • Prefer the Web Store when available; we also host a landing page and an unpacked zip for side-load or QA.
Download extension (.zip)

Set NEXT_PUBLIC_HASHTAGSPACE_CHROME_STORE_URL to show a "Chrome Web Store" button. The .zip above is the unpacked extension (Chrome → Extensions → Developer mode → Load unpacked).

Desktop

Native app (Mac & Windows)

Wrapped shell around this site—focused window, same login as the web. Read more about the desktop app.

Download installer
Local Render · Developer (NVIDIA GPU)

GIGI Local Render — M0 benchmark kit

The first step toward running the conversational-video avatar locally on your own GPU (so the per-call GPU cost goes to zero). This kit benchmarks whether a single modern NVIDIA card — e.g. an RTX 4090 — can render a talking-head avatar in real time. Grab it here on your Mac, move it to the GPU PC, and run it there.

  • Runs immediately with a synthetic baseline (no model download) to read your card's raw FPS, then wires in the real candidate models (SoulX-FlashTalk / JoyStreamer / LiveTalk).
  • Windows + CUDA 12.1 + PyTorch. Reports FPS, warm-start latency, and VRAM, with a 16 FPS / <1 s decision gate. See the README inside.
Download M0 kit (.zip)

Developer tool. Unzip, follow README.md, and run python bench.py --baseline on the GPU machine. Also install the #space Chrome extension on that PC (above) — the local render host will talk to it over Native Messaging in the next milestone.