Voice

Talk to Naseem, and let it talk back. Everything here is off until you turn it on, and everything runs on your Mac unless you choose otherwise.

Naseem speaks

• Read one reply. Every answer has a speaker button next to Copy. Click it and Naseem reads that reply; click again to stop. Code blocks and tables are skipped — they only make sense on screen.
• Read every reply. Turn on Reply aloud in Settings ▸ Voice, or click the speaker icon in the toolbar (it's the same switch). Naseem then reads each reply as it finishes. Choose what it reads under Read: the opening — the first sentences — or the whole reply.
• The right language, by itself. Naseem picks a voice that matches the reply, so an Arabic answer is read in Arabic and an English one in English. No setting needed.
• A word when you're away. Turn on Speak a summary when Naseem isn't in front and a run that finishes while you're in another app tells you what it did — one sentence. It never talks over you while you're reading.
• Stop at any time. Click the speaker button on the message, the toolbar icon, or Stop in Settings ▸ Voice. If you were dictating, the microphone comes straight back.

You speak

• Start. Click the microphone in the composer. The first time, macOS asks to allow the microphone and speech recognition — say yes to both.
• It keeps listening. The mic stays on until you click it again, so you never touch the keyboard between messages. What it hears lands in the composer for you to read; it never sends by itself unless you ask it to.
• Two ways to send, hands-free. Both are in Settings ▸ Voice ▸ Listening; use either or both.
• Say its name. End with the word Naseem (or نسيم) and the message goes; the name itself isn't sent. On by default. Turn it off if you say the name in your sentences.
• Just pause. With Send after a pause on, going quiet for a few seconds sends. The last second shows a countdown under the composer, and saying anything cancels it. Off by default — leave it off if you pause to think mid-sentence.
• Naseem waits its turn. After a message goes, the mic rests while Naseem answers (the line under the composer says so) and comes back on its own for the next message — after the reply is read out when Reply aloud is on, or as soon as it's written when it isn't. Naseem never dictates its own voice back into the composer.
• Where your voice goes. Under the composer, Naseem tells you whether your voice is recognised on this Mac or by Apple. That depends on the language you chose in Microphone language and what your Mac has installed; when it's on this Mac, your voice never leaves it.
• Typed and spoken mix. Anything already in the composer stays; dictation is added after it. macOS's own Dictation key works in the composer too.

Voices

• This Mac is the default engine: free, offline, in every language macOS speaks. Pick a specific voice or leave it on Best available for each language, and set the speed. Better voices can be downloaded in System Settings ▸ Accessibility ▸ Spoken Content.
• ElevenLabs, with your own key. (Pro) Switch Engine in Settings ▸ Voice and paste your ElevenLabs API key. Pick Flash for the fastest, cheapest speech or Multilingual when Arabic quality matters more than speed — both speak Arabic. The key stays in your Keychain and goes only to ElevenLabs; every sentence is billed to your ElevenLabs account, never to Naseem.
• One voice speaks every language. An ElevenLabs stock voice reads Arabic and English alike — the language follows the reply, not a setting. Pick the voice you like the sound of and leave it. This is the short answer for almost everyone: you do not need a second voice for Arabic.
• Browse by language is a filter. The picker above the Voice menu narrows the list and scopes the Library browser. It does not change the language Naseem speaks. Switching it to Arabic and back does nothing to your speech — but it does reorder the menu, so check the voice you end up on is still the one you meant.
• Choosing a voice. The Voice menu shows two groups: voices native to the language you're browsing, and Also speak … for the rest. Each is labelled with accent and gender — and voices you added yourself say so, since two voices can share a first name. A voice ID can always be pasted directly.
• Voices marked Library need a paid ElevenLabs plan. The Voice Library has voices native to a language — a real Arabic speaker rather than an English one reading Arabic — and Browse the Voice Library… adds one to your account. Listing and adding are free, but on a free ElevenLabs plan speaking with one is refused, and Naseem will say so under the reply instead of making a sound. Stock voices work on every plan.
• If a reply doesn't speak, the reason appears under it in orange — a voice your plan won't allow, a key ElevenLabs won't accept. Click the line to go straight to the setting that fixes it.
• Listing and adding voices need key permissions. When you create the key on elevenlabs.io (Developers ▸ API Keys), set Voices to Read and Write, and keep Text to Speech on. A key made without those permissions still speaks with the stock voices, but Naseem will tell you it cannot list or add voices.

Who it's for

• Anyone who thinks faster than they type, or is away from the keyboard: cooking, driving the Simulator, reading a printed page.
• People who can't use a keyboard all day. Naseem can be run by voice from the first word to the spoken reply.
• Language learners: dictate in the language you're learning, hear the reply in it, and switch the microphone language to match.