AI Tone Analyzer

About the AI Tone Analyzer

Paste an email, a chat message or any text and see how it is likely to come across: formal or informal, polite or blunt, confident or hesitant, friendly or cold, and how urgent, apologetic, angry or enthusiastic it sounds. The reading is shown for the whole text and sentence by sentence, with the exact phrases behind it marked in your text — so you can soften a demand, cut an apology or warm up a reply before you send it.

The basic reading updates as you type and uses fixed cues — words and phrases, punctuation, capitals, contractions, greetings and requests — with no AI and nothing sent anywhere. AI reading on this device re-reads every sentence with a small tone model that runs in your browser (downloaded once, after you agree). Explain with AI adds a short explanation in words, quoting the phrases that set the tone and suggesting changes.

It reads the text, not the person: it never guesses anyone’s personality, feelings or intentions.

Common uses

• Checking an email to a manager, a client or a teacher before you send it.
• Making a complaint firm without sounding rude, or an apology sincere without grovelling.
• Seeing whether a cover letter or a request reads confident or unsure.
• Reading a reply you received more calmly: whether it is really cold, or just short.

What each reading means

• Formal ↔ informal: register — greetings and sign-offs, contractions, slang, long or short words.
• Polite ↔ blunt: softened requests (“please”, “could you”, “would you mind”) against commands and demands (“send it now”, “you must”).
• Confident ↔ hesitant: firm statements (“we will”, “I recommend”) against hedges (“I think”, “maybe”, “just”, “I was wondering”).
• Friendly ↔ cold: warmth (“hope you are well”, “thanks so much”) against curt or distant phrases (“as per my last email”, “noted”) and several short sentences with no greeting, sign-off or thanks.
• Urgent, apologetic, angry, enthusiastic: how strongly the text reads that way, from words, exclamation marks and capitals. Capitals and “!!!” make a sentence’s own feeling stronger: “THANK YOU SO MUCH!!!” reads enthusiastic, “Send it now!!!” angry.

Each reading is a strength from 0 % to 100 % — how clearly the text reads that way — not a grade. A neutral text — plain facts — can sit near the middle of every scale.

Basic reading and AI reading

The basic reading is a fixed list of English cues (and a few Hindi and Hinglish ones such as kripya, jaldi and yaar). It is instant, works offline and explains itself — every reading comes from phrases you can see. A negation right before a cue cancels it (“not urgent”, “I’m not angry”), but the cues cannot see meaning that is not in their words, such as sarcasm.

The AI reading asks a zero-shot classification model, once per sentence, how well each tone label describes the sentence. It catches tone that is not in any list and reads other languages too (a multilingual model is used for them). It runs on your device after a one-time download and reads the first 25 sentences; for a longer text the result says which sentences it covers. The phrases marked in your text still come from the cues, so you can see what the words are doing.

What it does not do

It reads a text, not a person. It does not guess anyone’s personality, emotions, honesty or mental state, and it is not a way to monitor employees, students or anyone else. The same words can land differently depending on the relationship, the culture and what came before — treat every reading as a second opinion on your draft, not a verdict.

Limitations

• Tone readings are estimates. Context, relationship and culture change how a text is read, and sarcasm or irony is easy to miss.
• The basic reading’s cues are English, with a few Hindi and Hinglish words; for other languages use the AI reading.
• The AI reading needs a one-time model download, reads the first 25 sentences and takes a few seconds per sentence on slower devices.
• The AI explanation can be wrong; quotes it gives that are not in your text are left out.
• Texts up to about 3,000 words (20,000 characters) are read; the first 60 sentences are listed one by one.

Privacy

The basic reading runs in your browser with no AI and no network. The AI reading downloads a tone model from MySmartCoPilot once (its size is shown first) and runs it on your device. Only if you choose MySmartCoPilot’s cloud AI for the explanation is your text sent, with email addresses, phone numbers and ID numbers masked first; it is not stored.