Spin for today’s topic, research it for up to an hour, then talk about it for 1 to 10 minutes. Your audio never leaves this device.
Topics from: Everyday, or pick a category
Daily Talk gives you one topic each day — anything from the placebo effect to how GPS works — and a simple routine: read about it for up to an hour, then explain it out loud for 1 to 10 minutes. Explaining a topic in your own words shows you what you actually understand, and a short daily habit builds fluency over time.
Your talk is recorded and turned into text in your browser, by a speech model that runs on your device; the audio is never uploaded. You get the transcript with its length, word count and speaking pace, and every finished talk adds a day to your streak. Transcription is English for now.
Signed in, you can also ask for AI feedback once a day: a score and a note for structure, clarity, depth and examples, what to try next time, and claims worth checking, which it doesn’t mark right or wrong. On System Design and Database topics it also points out technical mistakes, with a correction, and the key points you missed. Only the transcript text is sent for that, never the audio.
A daily speaking-practice routine. You get one topic a day, read about it for up to an hour, then talk about it out loud for 1 to 10 minutes. Your talk is turned into text in your browser, so you can read back exactly what you said, with its length, word count, speaking pace and filler words. Signed in, you can also get AI feedback on it once a day.
Choose Everyday or one category, then spin for a topic from a full-screen picker. When you press Start, that topic is locked for the day and your research time begins: up to an hour, with links to Wikipedia (and Hindi Wikipedia where there is an article), Google and YouTube, and a notes box. Press “I’m ready” whenever you are — the hour is a maximum, not a target. Then record your talk any time before midnight; a talk you have already started can run past midnight. You can press Start until 11:45 pm; if you start late, research ends 10 minutes before midnight so there is time to talk.
Yes, before you press Start: reroll as often as you like (within your category, if you chose one). Once you start, the topic is locked until tomorrow. You won’t be given a topic you have already started until you have been through all the topics you are choosing from: all of Everyday’s, or all of that category’s.
Yes. Before you spin, choose Everyday (the general categories) or one category; technical categories only come up when you choose one. Spins and rerolls then stay in your choice, and the choice is remembered on this device. You can also switch category in the picker before you press Start.
English for now. Transcription is set up for English, so a talk in another language won’t come out as a usable transcript.
No. Your talk is recorded and turned into text by a speech model running in your browser, and the audio is never uploaded. Recordings, transcripts, research notes and your streak are stored in this browser. If you ask for AI feedback or a clean-up, only the transcript text is sent.
Transcription uses a speech model (Whisper base, about 80 MB) that runs on your device. It downloads from this site the first time you record — you can start talking while it loads — and your browser keeps it, so later talks don’t download it again. Clearing this site’s data from your browser removes it.
A day counts when you finish a talk: at least one minute long, stopped, and with its transcript saved. Today still counts until midnight, so the streak isn’t shown as lost in the morning; a day without a finished talk resets it to zero. The streak is kept in this browser, separately for each account.
Not to practise: you can spin, research, record, and read your transcript and stats without signing in. AI feedback needs sign-in. If you do sign in, Daily Talk keeps that account’s topic, talks and streak apart from the signed-out ones on this device. Nothing syncs between devices.
Yes, if you sign in. After a finished talk you can ask for AI feedback on your transcript: a score and a note each for structure, clarity, depth and examples, what worked, what to try next time, and a few claims worth checking, which it doesn’t mark right or wrong. For System Design and Database topics it also points out statements that look technically wrong, each with a correction, and which of the topic’s points to cover you left out. You can also have the transcript cleaned up first: punctuation and clearly mis-heard words only, shown before and after, and discarded if it changes too much.
Only text: your transcript and today’s topic, sent to an AI provider when you ask for feedback or clean-up. Your audio is never sent — it doesn’t leave your device.
Signed in, you get one free attempt each day, covering clean-up and feedback for that day’s talk. Signed out, AI feedback isn’t available. It never uses credits.
See more frequently asked questions