Speaking · Mouth
Speaking is RainbowSay’s session for practicing a language with your own voice. Hold space, say a sentence in the language you’re learning, let go — the app transcribes what you said, translates it back into your language and speaks it aloud, glosses every word you used, and gives you one correction. Then you ask why.
Animated demo of a Speaking session in push-to-talk mode: an English speaker practicing Spanish — drawn as a woman at her desk speaking into a microphone, next to the app. While the round microphone button is held, a level meter follows her voice and the learner’s Spanish sentence is written out as it was heard: “¿Podemos tener una mesa para dos cerca a la ventana?” On release the app speaks and shows what that meant in English — “Can we have a table for two near the window?” — so the learner can check it came across. The card then lists every word they used (Podemos, tener, una mesa, para dos, cerca, la ventana) and gives one correction: “cerca de la ventana” — “cerca” takes “de”, not “a” — with a more natural way to ask, “¿Nos da una mesa para dos?” The learner replays their own sentence with a word-by-word highlight, then asks in the chat, “Why «de» and not «a»?”, and gets a short explanation.
Animated demo of a Speaking session in push-to-talk mode: an English speaker practicing Spanish — drawn as a woman at her desk speaking into a microphone, next to the app. While the round microphone button is held, a level meter follows her voice and the learner’s Spanish sentence is written out as it was heard: “¿Podemos tener una mesa para dos cerca a la ventana?” On release the app speaks and shows what that meant in English — “Can we have a table for two near the window?” — so the learner can check it came across. The card then lists every word they used (Podemos, tener, una mesa, para dos, cerca, la ventana) and gives one correction: “cerca de la ventana” — “cerca” takes “de”, not “a” — with a more natural way to ask, “¿Nos da una mesa para dos?” The learner replays their own sentence with a word-by-word highlight, then asks in the chat, “Why «de» and not «a»?”, and gets a short explanation.
How it works
In the floating card, select Speaking. The first time, press Grant… so macOS lets the app use your microphone. Pick any microphone — built-in, USB, Bluetooth or a nearby iPhone — and watch its live level to see that you are heard.
Input language is the one you will speak — the language you are learning. Output language is the one you want it translated back into. Any pair of the 57 languages.
Press Start Speaking, hold the spacebar or the round microphone button, say your sentence, and let go. Prefer hands-free? Switch to Auto and the app hears where each sentence ends.
The card shows what the app heard you say, the translation back into your language — spoken aloud, so you know it meant what you intended — every word you used with its meaning, and one correction. Replay yourself, ask the chat why, and say the next one.
Speaking needs the Microphone permission, and Accessibility so the spacebar works as push-to-talk while another app has focus. It never records your screen, and the microphone is used only during a Speaking session.
01 · Push-to-talk or hands-free
Hold the spacebar — from any app — or the round microphone button, say your sentence, release. Nothing is sent unless you are holding, so you can think before you speak, and a noisy room never becomes a sentence. Or switch to Auto and simply talk.
Hold space, from anywhere. The key works while another app has focus. A short tap on the mic button latches it for people who can’t hold a key.
Nothing clipped. A sliver of audio from just before you pressed is included, so a word you started a beat early is not lost. A take ends by itself after 60 seconds.
Auto, for a rhythm. The app finds the end of each sentence and, once the translation has been spoken, listens again by itself: speak, listen, speak.
02 · Transcribed, then translated back
The card first shows your sentence written out exactly as it was heard — a quick check on whether you were understood. Then the translation back into your own language, spoken aloud. If it says what you meant, you said it right. If it doesn’t, you know precisely where.
As heard, not as hoped. The transcript is honest. A word that came out wrong shows up wrong, which is exactly what you need to see.
Spoken back to you. The translation plays aloud as its text lights up, in the voice and at the pace you choose.
Replay yourself. Your own sentence and the translation, side by side, with the word-by-word highlight.
03 · Every word you used
Your own sentence, glossed word by word in the order you said it: the verb form you reached for, the phrase that worked, the article you guessed. It is the vocabulary list of your own speech — the words you actually have, not the ones a course thinks you should.
Linked to your sentence. Point at a word on the card and its line lights up, and the other way round.
Phrases as units. “una mesa”, “para dos”: the card groups what belongs together, the way you would remember it.
04 · One correction
A patient coach doesn’t hand you a page of red ink. The card gives you the single most important fix in what you said, with the corrected phrase and often a more natural way to put it. Nothing to fix? It names the pattern you used best, so you know what to keep.
The mistake that matters. “cerca de, not cerca a” — one point, explained, instead of ten you won’t remember.
The natural version. How a native speaker would actually ask it, when yours was correct but stiff.
Encouragement is information too. When the sentence was right, the note tells you what you did well.
05 · Ask why
Why «de» and not «a»? Would the subjunctive be better here? How would you say this to a friend? Open the chat under the card and ask about your own sentence. The answer knows what you said, what you meant and what was corrected — and stays with the sentence in your history.
About your sentence. No pasting, no context to explain: the chat starts from the card.
Kept for later. Questions and answers are saved with the sentence, ready to re-read before the next attempt.
Who it’s for
Understanding is half of a language. Speaking is for the other half — wherever you happen to be practicing it.
Students
Teachers
Creators
Also worth knowing
A microphone hears your speakers. The app closes the microphone while it speaks or replays a clip, and for a moment after, so its own voice is never translated as your next sentence.
The picker lists every input with a live level bar. Set Noise reduction to Far mic for a laptop or conference microphone, Near mic for a headset or earbuds.
Choose and test your microphoneMute for a cough or a phone call. In Auto mode a long silence puts the session on standby; it wakes when you speak. In push-to-talk, waiting for a press is simply the normal state.
Describe your day, retell a scene from the show you are watching, answer the questions you expect at the interview — the guide has routines to start from.
Practice routinesQuestions
Something else? Write to [email protected] — a person answers.
Four things, on one card: what the app heard you say, written out in the language you spoke; the translation back into your own language, written and spoken aloud, so you can check it means what you intended; every word you used, with its meaning; and one note — the most important mistake in the sentence with the corrected phrase, or, when there is nothing to fix, the pattern you used best. You can replay your own sentence word by word and ask the chat why.
It does not grade pronunciation. What it shows is what was understood: the transcript is what the app heard, so a word that came out wrong shows up wrong — or as another word — in the transcript and in the translation. That is the most honest check there is on whether you were understood, and the replay lets you hear yourself next to the app’s voice.
Push-to-talk, the default: hold the spacebar — from any app — or the round microphone button while you speak, and release to translate. Nothing is sent unless you are holding. A short tap on the button latches it for people who cannot hold a key. Auto: just talk, and the app finds the end of each sentence the way a Listening session does; after the spoken translation finishes, the microphone re-arms by itself so you can keep a rhythm going.
No. The microphone would hear the app’s own spoken translation, so the app closes it while it speaks or replays a clip and for a brief moment afterwards while the sound dies away in the room. A headset does give the cleanest result, and the Near mic noise-reduction setting is made for one.
Any of the 57 languages RainbowSay works with — Spanish, French, German, Italian, Japanese, Korean, Chinese, Hindi, Arabic and many more — translated back into any other of them. The input language is the one you speak; the output language is the one you want to be understood in.
Any input macOS knows: the built-in microphone, a USB or Bluetooth headset, a lapel mic, an iPhone used as a microphone via Continuity. The picker shows a live level for each one while it is open, so you can see which one hears you, and Noise reduction is set per microphone type — far field for a laptop or a conference mic, near field for a headset.
Each sentence you say is saved with its audio clip in your study history on your Mac — that is what lets you replay yourself later — and you can delete any of it. While a session is running, the audio is processed by our AI provider, OpenAI, to transcribe and translate it. Nothing is uploaded to RainbowSay’s own servers but usage totals for billing.
Same card, different source. Listening hears a video or podcast and explains the sentences other people say; Speaking hears your microphone and explains the sentence you said — with one correction instead of a grammar note. Both share the chat, the replay and your study history, and one allowance pays for either.
Every new account gets one-time starter credit — no credit card — to spend on Speaking, Listening or Subtitles. When it is used up, you pick a plan in the app. Students, teachers and creators can email [email protected] for extra free credit first.
Download RainbowSay, hold space, and say the first sentence. One gentle correction at a time.
No credit card · audio only, never your screen · your history stays on your Mac
Also in RainbowSay
Every plan is one allowance, spent on any mix of the three. See how they fit together.
Transcribe and translate any show, video or podcast on your Mac, one sentence at a time — with the vocabulary and grammar explained.
Live translated subtitles over any video, call or your own voice — the original transcript plus up to three languages, in a transparent window.