Add "Dictation Mode" - Audio-only typing test for accessibility and transcription practice

[subhashkumar9661]

Is your feature request related to a problem?
Yes. Currently, all typing modes in Monkeytype are visual (reading words from the screen). However, in real-world scenarios like meetings, lectures, and transcription jobs, users need to type by listening. There is currently no modern platform that provides a dedicated "Audio-only" typing speed test.

Describe the solution you'd like
I propose adding a "Dictation Mode" (perhaps in Funbox or as a new mode) where:

The words on the screen are hidden or blurred initially.

The system uses the Web Speech API to dictate the words.

Users type what they hear, and the system tracks WPM and Accuracy based on the audio output.

Option to adjust the "Dictation Speed" (0.8x, 1.0x, 1.2x).

Describe alternatives you've considered
I have checked existing websites like Sense-Lang or GoTranscript, but they are outdated and lack the modern UI/UX and customization that Monkeytype offers.

Additional context
This feature would significantly improve accessibility for visually impaired users and provide a unique training tool for professional transcribers. I am a Computer Science student and I'm very interested in contributing to this feature!

[fehmer]

a very basic variant of this is available as funbox "tts"

[fehmer]

Sounds very cool. could you join the discord in #development

[abhinavgautam01]

Hey, this idea sounds really interesting ...
Is anyone actively working on the steady-stream dictation part right now? I’d love to help contribute, especially around improving the audio flow / UX for a proper audio-to-type experience.

[KitKat20255]

In my case learning to type on a new layout was a horrible experience using quotes, i noticed that my speed on quotes went up to 40 right away, but my speed from writing my own thoughts was 20 at best, so i realized that if you want to get good at writing your ideas or dictation, you have to practice that specifically, so dictation it's the best tool for me. I personally start a youtube video in the background, and i try to write in the ZEN mode as much as i can remember, sometimes i make up words or sentences to cover no talking parts, that's the most comfortable practice mode for me right now.

[KitKat20255]

To add to this suggestion which was merged with this one #7929 . You can also have a 3rd parameter or some AI stuff that would dictate the speed of read ahead. When a quote it's added in the list, the script can identify the fast and slow parts of the quote depending on the first performances of other typists, and increase the pace where the words are very short and common and it's expected a spike in the speed of the typist. It can slowdown on the hard words, it can detect dots and commas in the quote and wait for the words read ahead to decrease slightly before starting a new sentence. In case it's hard to understand what i mean, a sentence like: "World war 3 it's coming and we are expecting that the global damage will be significant" for a super fast typist the script can identify breaks in a sentence like commas, or changes of the flow like the word "and" "then" "however" "meanwhile" "but" and it will read parts between breaks at a speed above their average speed. So a 150 wpm typist would hear: "World war 3 it's coming: , once they reach the letter "r" from "war" which is exactly 15 characters from the end of the sequence read, the next sequence it's read, or part of it if it's too long. In this case once it types that "r", the tts will say "and we are expecting" until the next break in the flow, once the typist types "e" from "we" (15 char before end) the tts will read "that the global damage" (ai can detect that words global and damage are connected words in this sentence and read them together), and again once it reaches "e" from "the" which is 15 letters before end, will finish the sentence to the dot, it's not a great flow to have tts dictate words like "it's" and take a break, a sentence should be broken down into parts easy to digest and understand the meaning of it. However if parts of the quote are consistently hard for many typists the trigger character can be decreased (in this case lower than 15), or increased for the super easy parts by a percent of their average speed. I think that dictation mode should be the main attraction of this entire website tbh, not just a "fun mode", but the main improvement tool.

[KitKat20255]

I think this is a good starting compromise if it's too difficult and ambitious ^^ Get one of the AI's to break down quotes like this, and the resulted text will be read by the TTS in sequences, and the next sequence will be triggered depending on the average speed of the typist, 10 characters before the end of this one, or 20 and so on, maybe even allow in settings to manually select a percentage/fixed increase based to your own attention span. Also ignore the "mistakes" that the user might type from using punctuation, but include them in the speed calculation. So we don't get penalized for using punctuation and capitals where we feel it makes sense.

(also when pressing the omg i forgot what to type key/shortcut) the current word that's being typed wrong should be deleted if exists, then the tts will first read the remaining of the current sequence, and start the next one when appropriate.