After reaching level 60 I’ve been trying to spend more time on conversation practice (mostly voice but some text too). I’ve been spending a lot of money on iTalki lately and started looking for alternatives.
Specifically I was looking for a compelling voice chatbot.
I didn’t really like anything I found so I decided to try to build one myself. I’ve been enjoying using it and thought I’d clean it up and make it available to others.
On the first screen, I clicked on the green “Maybe start by saying…” (and it accepted my voice in blue) but then after it responded to me in gray I tried clicking in the entry field at the bottom, or by clicking on the microphone icon, but could not figure out how to get it to listen to me.
Clicking on mic icon gets ‘speech recognition service not allowed’ (can’t capture message in screen shot)
Thanks! Pretty much anything that goes wrong at this stage is on me
The “speech recognition service not allowed” error means the site can’t get access to your mic. Usually it’s because you didn’t grant it permission. I’m on windows atm but if I recall right there is a Safari > Settings for this website where you can control mic/camera access.
I think the speech recognition api I’m using is good for Safari 14.1+
I found a setting to turn it on - it will require restarting safari to take effect, which I am not ready to do yet.
How should it have worked? Did I need to click in the entry field at the bottom of the page to speak, or click on the microphone icon, or something else?
When your mic is accessible the Mic button is green. You click it, speak, and your words pop up in the text field at the bottom.
If there are imperfections in the words it heard you can edit before sending (don’t cheat and type the whole msg, the point is to speak )
If you’d rather just have a back and forth, click on the little settings icon and switch on “Conversation Mode.” That automatically turns on your mic every time Pera’s done talking.
If you’d like your response sent automatically when the mic turns off, switch on the Auto-send option.
If you’re on a mobile device and want to hold down the mic to talk (like Messenger, Whatsapp, etc.), then turn on Walkie-Talkie mode.
Speech recognition This has been completely overhauled. Error rates should be down at least an order of magnitude. Pera also now recognizes her own name more accurately (instead of hearing 寺さん and the like)
Translation and Transliteration Both translation to English and romaji are working well. Translations are now much more accurate. Let me know if would like support for other target languages. Furigana also on the way.
Atmosphere Added option for ambient sounds to match scenario. If taking a walk in the park with Pera, immerse yourself in the sounds of a peaceful Japanese garden.
Dark mode Since I spend a considerable time practicing with Pera I’ve switched to dark mode as default to reduce eye strain.
Patience You can now control how long Pera waits for you to speak. For a snappy, quick conversation keep Patience low. For beginners and slow speakers, you can increase it.
Support Added a help page to answer some questions that came up during early testing.
Coming up next:
Stats/Dashboard Your speech stats and visualizations so you can monitor progress.
Corrections Pera will correct your grammatical/spelling mistakes
Visuals Pera will be able to generate visuals to help explain terms and concepts
Romaji Default If you can’t read Kana/Kanji Pera can use Romaji by default.
Let me know what would make it more helpful for you. Also, more users recently has meant more load on the backend and more costs so I’ve limited messages per chat. You can remove this with a small upgrade if you wish.
More updates to make Pera more beginner friendly based on feedback:
Onboarding Tour: you’re now shown a tour on how to use Pera
Beginner by Default: Now Pera assumes you’re a beginner by default. You can change your proficiency in the Settings. Pera behaves differently based on your level.
Automatic Romaji for beginners.
Automatic Translations for beginners
Hints to start the conversation for beginners
Feedback Button: that goes straight to my inbox
Other Additions: pagination, cleaner layout, more control over settings
Take a look, appreciate all feedback → heypera.com
Reminder: Pera is a Japanese conversation partner. I built Pera after hitting 60 and deciding to focus more on speaking
The goal is WK for kanji, Bunpro for grammar, and now Pera for speaking.
My goals with Pera:
Always available conversation partner to talk about anything.
Must have a memory, that’s able to learn about you over time, so conversations develop realistically
Absolute beginner friendly. That means the 3 T’s: transcription, translation, and transliteration (romaji). It also means that when you’re stuck, partner should suggest what to say next, how to pronounce it, and what it means.
Scenario specific conversations. E.g. how do you talk when you visit a doctor, get your hair cut, etc. Some of that language doesn’t come up naturally in an open-ended conversation.
Corrections when you screw up.
Much of this is based on my experience on what worked with paid conversation partners/tutors who I’ve spent thousands on.
Coming up next:
Focus Mode: Pick a grammar points, vocabulary, kanji, etc to practice in a conversation
You can use it for free but with limits because the compute behind it is pricey (though not as pricey as an hour with a tutor - so upgrade if you want).
Try it out, lmk what you think and if you have any requests/feedback
I’ve been pasting feedback here (feel free to do the same): Feature Requests | HeyPera
)
