Read Japanese text aloud with AI
AI reads the Japanese text you enter aloud, and you can save it as an audio file. The speech synthesis AI Irodori-TTS runs inside your browser, so the text is never sent anywhere. No app to install and no sign-up. Use Chrome or Edge on a PC with a GPU (WebGPU).
This tool reads Japanese text. Text in other languages is not read correctly.
Writing numbers as digits, as in “10部”, reduces misreadings. Words that are misread can be added to the “Pronunciation dictionary”.
AI assistance also runs in your browser, and the text and your request are never sent anywhere.
Voice and quality
If you leave this empty, a voice description that suits the text is generated.
The dictionary is stored in this browser and never sent anywhere. Before reading aloud, words in the text are replaced with their readings from the dictionary.
Output settings
AI models stored in this browser
How to use
- 1
Enter the text
Type or paste the Japanese text to read aloud. Writing numbers as digits, as in “10部”, reduces misreadings. Pressing an emoji button inserts an emotion or speaking style at the cursor.
- 2
Choose the voice and quality
Choose a voice from the list, or choose “Write your own voice description” and describe it in Japanese. The default quality is “Standard”.
- 3
Read aloud and save
Press “Read aloud”. The first run downloads the AI model (about 1.7 GB). Remake any misread part with “Regenerate”, then press “Save” to get an audio file.
How it works, and setting the voice and emotion
Choose a voice from the list, or describe one in Japanese in “Voice description”, such as “落ち着いた低めの声の男性” (a calm man with a fairly low voice). Emoji in the text set the emotion or speaking style at that point (whispering, pauses, laughter and so on). Long text is split and generated in order, and the rest is generated to match the voice of the first part, so the same voice reads it to the end. The AI is the speech synthesis model Irodori-TTS (MIT license). The AI model (about 1.7 GB) is downloaded from Hugging Face only the first time and stored on your device for later use.
You are free to use the audio you create. No credit is required. Using it to impersonate a real person, or to make people believe a real person said something, is prohibited. Because the AI creates the voice anew each time, it may happen to sound like a real person.
Frequently asked questions
Is my text sent to a server?
No. Text to speech runs in the browser on your device, and the text you enter, the voice description and the pronunciation dictionary are never sent anywhere. The AI model is downloaded from Hugging Face the first time and stored in your browser. No sign-up is needed. Processing statistics used to fix bugs (device specs, the approximate number of characters and sentences, error types and so on; never the text itself or the voice description) are sent to this site's server, and you can turn this off with the switch under “Help us improve the service” at the bottom of the page.
Does it work on iPhone or other phones?
No, it does not work on iPhone or iPad. Once loaded, the text-to-speech AI uses about 3 GB of memory for its calculations. On iPhone and iPad, the operating system limits how much memory a single page can use, and on the iPhone 15 Pro Max (8 GB of memory) we tested, the page went over that limit after the AI model was downloaded and reloaded automatically before any speech was generated. So that the roughly 1.7 GB download is not wasted, the tool stops on iPhone and iPad before the download starts. Use Chrome or Edge on a PC with a GPU (WebGPU). It also runs on Android and in Safari on a Mac, but it may stop partway on some devices.
Can it clone a voice (copy my own voice or a voice I like)?
No. So that it cannot be used to copy someone’s voice without permission or to impersonate a real person, we deliberately left out any feature that loads a recording or an audio file to imitate a voice (voice cloning). You specify the voice with a written “Voice description”. A voice you like can be saved in this browser with “Save this voice” and used to read any text in the same voice. When you export a saved voice to a file, only how the voice was made is recorded. On import, this tool recreates the voice from that, so a voice made from a recording cannot be brought in.
Can it read long text?
Yes. The text is split at sentence breaks, read in order and joined into one audio file with short silences in between. There is no character limit. On a PC we tested (the integrated GPU of a Ryzen 7 8745H), “Standard” makes 1 minute of audio in about 30 seconds. “High quality” takes about 6 times as long.
Help us improve the service
If a conversion fails, sending us the circumstances helps us track down the cause. Knowing which devices and formats fail most often, and how long processing takes, lets us revise the supported formats and fix the bugs behind them. If you don't mind, leaving this on is a real help.
What is sent
The tool you used and the settings you chose (output format, quality, etc.), your device specs (GPU, memory, core count), the input format with a rough size (for text to speech: the approximate number of characters and sentences, how many parts the text was generated in, and how long the AI model took to load), how many files succeeded or failed and how long it took, and the error type and details when something fails.
What is never sent
The files themselves, file names, text, subtitles or prompts you add to a video, the text to be read aloud, voice descriptions, the pronunciation dictionary and saved voices, and IP addresses. The data goes only to this site's own server, never to a third-party service.