Dictate a short message (3-5 sentences), pronouncing punctuation marks and structure in your voice, and then give the command to clean. Return JSON {"raw_dictation":"...","cleanup_instruction":"...","result":"..."}: raw dictation with voice commands, instructions for combing hair and final clean text.
Dictation is the translation of voice into text in real time. The model hears sound, predicts words based on the sound and meaning of neighboring words, so coherent speech is recognized more accurately than jagged fragments. She doesn’t always put punctuation marks herself - they can be pronounced out loud: “comma”, “full stop”, “new line”, “paragraph”. This is the first lever: dictate the structure out loud, and the output will be a finished text, not a continuous one. The second lever is more important - don’t try to dictate purely. It’s faster to say the thought as it is, with reservations and repetitions, and then with one command ask the AI: “remove the repetitions, comb your hair, leave the meaning.” The voice gives volume and speed, the model gives purity. Insider: dictate in long chunks rather than one word at a time - context helps recognition and accuracy increases. A typical mistake is to dictate in a noisy place with jumping volume and then be surprised by the mistakes; Even volume and a close mic do more than any setting. And always double-check numbers and names - the model confuses them most often.
Unlock access to submit solutions for instant AI review.