AI that reads its answers aloud, free and with no sign-up

When voice is better than reading and typing

Long answers are easier to take in while you cook, walk or rest your eyes. Read aloud lets you listen to a recipe, an explanation or a story instead of scrolling through it. Dictation helps when a question is easier to say than to type: on a phone, when you are thinking out loud, or when typing is slow or tiring for you.

Both work in a normal chat, so you can mix them freely: speak a question, read the answer, then have the next one read to you.

Use voice in three steps

Step 1: Open nokeep in Chat mode. There is no sign-up screen. To dictate, press the microphone next to the message box and allow your browser to use it.

Step 2: Speak your message, then press the microphone again to stop. Recording stops on its own after 90 seconds. Your words appear in the message box, where you can fix anything before you send.

Step 3: When the reply arrives, press Read aloud, the speaker button under it. The first words play within a few seconds while the rest is prepared. Press it again to stop.

What is free, and what costs credits

Free use includes 100 voice uses a day, shared between dictation and read aloud, with no account. Each dictated clip is one use. A long reply is read in parts, a short opening part and then parts of about 2,000 characters, and each part is one use, so a very long answer can take a few.

With a plan or credits, voice is billed by length rather than counted. Read aloud is about 57 credits per 1,000 characters, about 5.7 cents, and dictation is about 23 credits a minute, about 2.3 cents, where 1,000 credits are one dollar. On a plan these come out of the plan’s monthly credits.

A voice call with a character uses the same voice allowance: about 10 minutes of a typical call a day for free, then about 50 credits a minute with credits. A quiet call costs less, because only what is said and heard is counted.

The free allowance resets each day and is the same for everyone who uses nokeep without paying.

What happens to your voice

When you dictate, the browser records a short clip and sends it to nokeep’s server, which passes it to a transcription service under zero-data-retention terms and returns the text. Nothing is written to disk on the way, and the recording is not kept by nokeep. Read aloud works the same way in the other direction: the reply’s text is turned into audio, streamed to you and not stored.

Your chats, including the dictated text, are saved in your browser rather than on our servers. nokeep keeps no voice samples and does not train on them. If voice is ever briefly unavailable, it is because its privacy terms are being re-checked, and it says so rather than switching to a different service.

Limits

In a normal chat there is one reading voice and no choice of speed. Characters are different: each one reads in a voice of their own, chosen when they are made and changeable from Voice on their page. Code blocks are skipped and announced as “code omitted” rather than read character by character.

Dictation needs a microphone and a browser that allows recording, and each clip is at most 90 seconds. For a long message, dictate in parts. Like any transcription, it can mishear names and unusual words, so glance at the text before sending.

In a normal chat this is not a live conversation: you speak, send, and listen, one turn at a time. For a hands-free conversation, call a character from their page. Calls are in beta: you talk, pause, and they answer out loud in their own voice, and the whole call stays in the chat as text.