feat: add terminal-tts demo

[luke-speechify]

Pairs with the upcoming speechify.ai post "Text-to-speech from your terminal: a CLI demo with the Speechify API".

What this demo does

A terminal-styled web playground for Speechify text-to-speech. The page renders a monospace "terminal" where you type a command like:

speechify say "hello world" --voice geffen_32 --model simba-3.2

Hitting enter parses the quoted text and --voice / --model flags client-side, POSTs { text, voiceId, model } to a server route, plays the returned audio, and prints shell-style output. Also supports voices, help, clear, and Up/Down command history.

The original idea — reading text aloud straight from the command line — can't be hosted as a shared page, so this is a hostable adaptation. The real CLI ships alongside it in cli/say.mjs: a standalone, dependency-free Node script (native fetch + node:fs) that calls POST https://api.speechify.ai/v1/audio/speech with your env key and writes an MP3 (or pipes to a player). Documented in the README under "Run it as a real CLI".

Key handling and gating

• Server-side key: all synthesis in the web app goes through app/api/say/route.ts (Node runtime, client.audio.speech, returns base64 MP3). SPEECHIFY_API_KEY never reaches the browser; @speechify/api is marked server-external.
• Turnstile: the widget renders near the terminal, and every gated fetch attaches the x-turnstile-token header. Fails open locally when TURNSTILE_SECRET_KEY is unset.
• The cli/say.mjs script reads the key from your own shell environment — it runs on your machine, not in a browser.

Hosting registration

• vercel.json — added a terminal-tts service + a /terminal-tts/:path* rewrite (before the / and catch-all rewrites).
• pnpm-workspace.yaml — added demos/terminal-tts.
• pnpm generate — regenerated the README table and site manifest.
• Dev port 8769 (unique across demos).

Verification

• pnpm --filter terminal-tts build succeeds (build + typecheck ✅).
• Live API call ✅ — ran the shipped cli/say.mjs against the real endpoint with the default geffen_32 / simba-3.2; got a valid 44 KB MP3 back (the web route sends the identical request body via the SDK).

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
speechify-api-demos	Ready	Preview	Aug 21, 2026 12:17am

[luke-speechify]

Demo queue review — terminal-tts:

• CLI matches the DRG-374 thesis: dependency-free cli/say.mjs calling POST https://api.speechify.ai/v1/audio/speech with model simba-3.2, key via SPEECHIFY_API_KEY. Endpoint, voice and model defaults match the other TypeScript demos and the docs.
• Key stays server-side: the Next.js playground proxies through route handlers; the CLI reads .env locally. No client key exposure.
• demo.json (order 205) + pnpm-workspace.yaml + vercel.json + regenerated README/index are all in this PR — hosted signal and generated outputs in sync.

Not approved (leaving approval to Luke per merge policy). One note: this is the canonical CLI demo for the Text-to-speech from your terminal post; a duplicate cli-tts PR was opened and withdrawn.