Cloud Speech

Turn highlighted text on any web page into natural speech with your own cloud TTS account: Amazon Polly, Azure Speech, Google Cloud TTS, OpenAI, or any OpenAI-compatible server, from one extension. Setup guides, pricing notes, and troubleshooting are on the website.

Features

• Connect Amazon Polly, Azure Speech, Google Cloud TTS, OpenAI, and/or any OpenAI-compatible server with your own credentials; every voice appears in one searchable picker, tagged by provider
• Preview any voice before selecting it, and star your favorites
• Read aloud from the context menu, a keyboard shortcut (Ctrl/Cmd+Shift+S), or the popup sandbox, with playback controls and live speed adjustment
• Download any selection as an MP3 file (Ctrl/Cmd+Shift+E)
• Adjust speed, pitch, volume, and speaking style where the selected voice supports them
• Use the interface in English, Chinese (Simplified and Traditional), or Hindi
• Credentials stay in your browser (Chrome sync optional), and text is sent only to the provider you chose

Store listings

One Chrome build goes to the Chrome Web Store; a separate Firefox build goes to addons.mozilla.org.

Listing	Store	Who gets it
Cloud Speech (the Polly for Chrome listing, renamed in place)	Chrome Web Store	new installs; former Polly users received it as a normal update
Cloud Speech	addons.mozilla.org	the Firefox build

Development

• Stack: WXT, React 19, TypeScript (strict), Tailwind CSS v4, Bun workspaces.
• apps/extension is the extension; apps/web is the Astro website. The scripts below run from the repo root.

bun install            # install dependencies
bun run dev            # extension dev with HMR (opens Chrome) + website on localhost:5173
bun run dev:extension  # the extension alone, with interactive WXT keys
bun run dev:web        # the website alone
bun run build          # check + all builds: chrome, firefox, web (browser builds also zip)
bun run build:chrome   # Chrome build + store zip -> apps/extension/.output/chrome-mv3
bun run build:firefox  # Firefox build + store zip -> apps/extension/.output/firefox-mv3
bun run build:web      # website -> apps/web/dist
bun run check:size     # size-limit budget on the built content script
bun run typecheck      # tsc --noEmit (strict, both apps)
bun run check          # biome lint + format (check:fix auto-fixes)
bun run test           # vitest, both build targets (chrome + firefox)
bun run test:coverage  # vitest with coverage thresholds
bun run lint:firefox   # Mozilla's addons-linter on the Firefox build
bun run verify:zips    # manifest smoke on the emitted store zips
bun run test:e2e       # Playwright popup smoke against the built extension (one-time: bunx playwright install chromium)
bun run test:e2e:firefox  # the same smoke in a stock Firefox through Selenium (needs a Firefox on PATH)

• Chrome: load apps/extension/.output/chrome-mv3/ unpacked from chrome://extensions (Developer mode).
• Firefox: bun run --cwd apps/extension dev:firefox runs it in a temporary profile through web-ext.
• Architecture rules and the provider contract: AGENTS.md.

Rebuilding the Firefox store package from source (AMO reviewers):

1. Install Bun at the version in .bun-version.
2. bun install --frozen-lockfile
3. bun run --cwd apps/extension build:firefox; the zip lands in apps/extension/.output/.

Contributing

• PR titles are Conventional Commits; CI gates on the all-green check.
• Before opening a PR, run what .github/workflows/checks.yml runs, all from the block above.
• Account-wide conventions: the contributing guide.

Security

Report vulnerabilities privately through the repository's Security tab (Report a vulnerability), never in an issue; the security policy has the details. In scope:

• Every listing ships the same source at the same version, so a report against one applies to all.
• API credentials live in chrome.storage (sync by default, local when the sync toggle is off). Anything that exfiltrates, logs, or leaks them is high severity.
• Selected text goes only to the provider the user configured, straight from the browser, with no intermediary servers or analytics. Any other destination is a bug.
• The content script runs on all pages (<all_urls>) for error toasts; selected text is read on demand through scripting.executeScript. Injection or privilege escalation in either path is in scope.

Support

• Bug reports and feature requests: GitHub Issues
• Website: vivswan.github.io/cloud-speech

License

• Individual and Small Organization License 1.1.0; see LICENSE for the terms and its summary table.
• Individuals may use it for any purpose they choose for themselves; organizations under 100 people and 10,000,000 USD yearly income may use it; anything beyond that needs the licensor's permission.
• Releases through v1.0.8 were published under the MIT License and remain available under it.