This blog post was brought to you by the generosity of AJATT's patrons!

If you would like to support the continuing production of AJATT content, please consider making a monthly donation through Patreon.

Right there ↑ . Go on. Click on it. Patrons get goodies like early access to content (days, weeks, months and even YEARS before everyone else), mutlimedia stuff and other goodies!

Chinese Project Notes 10: Big Developments (Anki, Text-To-Speech, Cantonese, Victory Calendar)

うっす. So, it’s been a while since I posted one of these, but anyway here we are all the way at #10. I’m going to try to keep this one short (I’m not just saying that!) because these suuuuuuuuuuuck to edit afterwards. Actually, they’re a lot better since I started using Microsoft Word a word processor from a certain major software manufacturer. But still, I’m tired of ending up with sentences like “you nee’ a use SRS-lah. I’s so simpo!” in long posts. OK, here we go!

A lot has changed. A lot. And I haven’t been telling you jack. Not because I’m a bad person, but because I don’t like to talk about things I’m not sure about. Because, generally, one of two things happens when you do this:

a. People get too excited, try it, but if it doesn’t work (since it wasn’t fully tested), then they feel bad, and maybe they tell you that you suck.


b. People shoot you down before you’ve even tried it, and (if you’re delicate like me) it kills your will to try, and we all know that not trying is the source of all failure. But I digress.

Sensitive KhatzumotoThis “not talking until the doing has been done” thing is one of the main reasons why I didn’t put up AllJapaneseAllTheTime.Com at the start of my Japanese journey. I left it till the end, when I had nothing left to prove as such, and any barbs directed at my person, real or imagined, would be functionally useless, since they cannot negate the simple fact that I have near-native Japanese ability now. You know, kind of like how only people who don’t have money are hurt by people thinking they don’t have money? Or something to that effect…I’m sure you understand what it’s like – the Internet is full of the most negative, demoralizing, borderline-to-overtly racist crap when it comes to East Asian languages, and normal, sensitive (see Fig. 1) people are easily harmed by it.

By the way, the other main reason is that making a website used to be annoying. Blogs have been around for a while, but I honestly thought that blogs were just for keeping diaries for a limited audience because that’s all that people used to do with them. That is, until I saw someone using a web log other than for logging, with articles actually written to be read by non-insiders, and that changed the game for me. Speaking of logging, Momoko encouraged me to keep a log of my Cantonese progress, even if I don’t actually post on it for a while. I am more or less doing that.

But you didn’t come here to hear that kind of beanbag philosophy (“dewd, like, isn’t it amazing how..”), back to the article.

Crap…what was I gonna say. OK, first stop is Anki and Text-to-Speech (TTS).

Text-to-Speech (and Anki)

In Chinese Project Notes # 8, I discussed changes I had made to my SRS entry format. Based on the effects of those changes, I have made even more alterations. Some I will discuss in this article, some may have to wait for later; there’s seriously that much going on.

First, why did I make these alterations? Well, I discovered that while the Chinese Project Notes # 8 changes were definitely a step forward for my handwriting – I can produce hanzi/kanji from memory with great speed and accuracy and exactly when I want them – the changes have not (yet?) given me the aural benefits that I had expected. My Chinese writing advanced to pwnage level, but my listening comprehension was not being all that it could be.

To the chase I am cutting. My calculations indicate that at this time it would not be economical to add free sound support for everyone on KhatzuMemo. Plus, Anki is a really good SRS, so why not try it out, right? That’s what I did. After tons of pride-swallowing, trial, and error, my Cantonese (and some Mandarin) SRS items essentially consist of:

  • Question:
[Audio of sentence]
  • Answer:
[Text of sentence: this is what you have to write out, given the audio] [Dictionary definitions, as necessary] [Translation of sentence, if necessary] [Phonetic reading for clarification, if necessary]

Here’s an example:

  • Question:

[Audio of sentence]

  • Answer:

你去邊? [Text of sentence: this is what you have to write out, given the audio]

你去哪裡? [Translation of sentence, if necessary]

Néih heui bīn[Phonetic reading for clarification, if necessary]

The process is basically that I am both chorusing (or parroting, or whatever) and taking dictation at the same time. I think dictation is one of the best language-learning exercises out there in that you are connecting the verbal and written parts of a language, something that a lot of people fail to do. It’s a hybrid input-output affair that puts almost all the skills that matter on the line – you have to understand what’s being said, and you have to know how to write it out exactly correctly. Chorusing, or what I am calling chorusing, is really good, too — listening to (native) speech and imitating it. Step-by-step it goes like this:

1. Play audio (as many times as necessary).

2. Say audio.

3. Write down text, based on audio (audio may be repeated).

4. Compare my text to the correct answer.

Where do you get the audio? I use text-to-speech (TTS) software. It set me back a bit, but I like to think of it as an educational expense. The TTS software I got comes in two parts – a reader, and voices. As far as I know, you need both. My reader and voices are:

  • TextAloud – the reader. It does cool things like managing text and converting it to MP3. I believe it comes with a basic, default English voice, but good voices and voices in other languages need to be purchased separately. There is a free trial version of TextAloud available here.
  • Voices. I use Lily for Mandarin, Sin-Ji for Cantonese and Misaki for Japanese. I chose female voices because I found them easier to understand. Maybe it’s a high-frequency thing? Or maybe it’s just my imagination – I don’t actually know for sure. Currently, I only use the Japanese one for reading me long articles, like the ones from this site.

TTS has been around a relatively long time. Why am I only now getting into it? Well, it used to suck; it was a running joke. TTS is much better now than it was 5 years ago, and while the voices are not yet perfectly human, if you’re a beginner, they’re almost certainly much closer to perfection (accurate pronunciation) than your voice is in your target language, which is what counts. The Japanese voices are especially blowing me away [audio sample of the first paragraph of this article].

Webcam KhatzumotoThere is also something special about the nature of Chinese that drove me to TTS. Other than Bopomofo/注音符號, there are no satisfactory phonetic systems for representing Chinese. By “satisfactory”, I mean “consistent, easy-to-understand, and will lead to native-like pronunciation if followed”. Pinyin sucks. Jyutping sucks even harder. Yale Mandarin is decent. Yale Cantonese is an improvement over Jyutping but still not all the way there. I needed to know how to pronounce Cantonese without, like, balancing an equation every two seconds (because that’s what tone numbers turn life into). The tone markers had no meaning to me – I could not differentiate them – until I actually heard a lot of Cantonese. I needed to focus on what Cantonese sounds like, because that’s what matters, not some trainwreck of a Romanization system. This is what led me in the direction of TTS. The results are good so far – one Cantonese speaker from Hong Kong on Skype accused me of lying about not being Chinese, despite my insistence that “it’s not that good…yet”, so I had to borrow a friend’s webcam (see Fig. 2), and then the Skype guy made me undress. It just goes to show that watching and/or listening to Cantonese dubs of American cartoons 18 hours a day doesn’t not have an effect. And, yes, I do randomly find Cantonese speakers on Skype to talk to. I learn a lot from them if I shut up. Skype chat records are automatically saved, so you can go back later and sentence-pick, and also to absorb the corrections you no doubt asked for.

One annoying problem with the Chinese TTS voices I use is that they cannot pronounce certain characters correctly or at all, especially ones used in written colloquial Cantonese, even some commonly used ones. Not only that, but they have no “learning” ability – you can’t “teach” (customize) them to pronounce certain things correctly. Misaki does have such ability; she can even be “taught” intonation…I look forward to a customizable Cantonese voice. At any rate, TTS is still a great tool, and I imagine many people could benefit from using it.

Note – you could try just cutting sound samples by yourself instead of using TTS. I have tried this; it’s good but it has its limitations – it takes time to do and it obviously won’t have the same vocabulary range as TTS. I primarily use TTS, but a mix of TTS and manual sound-clipping seems like it would be a great combination.

Cantonese: What’s Up With That?

As you may be aware, Cantonese has been “on my radar” for quite some time. When I made the decision to learn it, I was already focusing on learning Mandarin. The reasonable thing to do, and what I initially chose to do, was to continue doing Mandarin until my Mandarin got really, really good.

So I started building a Mandarin immersion environment. That involved getting Mandarin dubs of my favorite American cartoons — stuff like 蝙蝠俠/Batman, 飛天少女驚/Powerpuff Girls, almost all the Disney/Pixar movies. As it turns out, almost all of these DVDs had a Cantonese track as well. Occasionally I would switch to the Cantonese track for laughs — it sounded so funny!

Anyway, this “funny-sounding” language or dialect started to grow on me. The Bruce Lee effect and the fact that (until recently) the Chinese that most non-Chinese people heard was in fact Cantonese, certainly played a part. Cantonese is even more “magical”, more BS-ed about, more Orientalized, more feared, more hyped than Japanese; this, I am sure, tickles my reverse-BS glands.

So it got to the point that I was just trying to “get through” Mandarin in order to get to what I really wanted to do – Cantonese…After much, much, much, deliberation and gnashing of teeth, I decided to go all Cantonese all the time; Momoko had gotten fed up of hearing me whine and worry compare and contrast. I continue to learn token amounts of Mandarin out of a feeling of necessity, no, duty, even. But I do Cantonese out of love and therefore Cantonese gets all my time now. If Mandarin and Cantonese are in danger of drowning, and I can only save one, Cantonese gets saved every time. There is so much Cantonese playing in my house that Momoko sometimes randomly says things like “開開心心”/heppy, whether or not she understands them. Repetition will do that.

Momoko randomly speaking Cantonese

Victory Calendar

Everyone who reads this site is incredibly good-looking and positive. And that helps. In fact, most of my fears and doubts are self-induced. But anyway, to keep me from sinking into fear, doubt and I-can’t-do-this-ism, I have made myself what I call a “Victory Calendar”. Wait, before I tell you about the calendar, let me just say this. I finally understand the sheer disbelief that I sometimes read from people who read this site. Because the method explained on these pages is so simpo. Just DO it. It’s THERE. You CAN. It’s so simpo that it would seem that anyone could do it, right? And anyone can. But if it’s so simple, why isn’t everyone doing it? Why are there people who have been living in Japan for 20 years and can’t even read hearmegana? Can’t even write one kanji?

Because, it’s just like Jim “the Rohnster” Rohn said – “the things that are easy to do, are easy not to do”. It is just as easy to eat fruit as to eat a candy bar. Just as easy to watch Powerpuff Girls in Cantonese as to…not watch Powerpuff Girls in Cantonese. What the Rohnster is saying is that the results, the achievements (or lack thereof) of our lives are the sum total of tiny, “insignificant” decisions. “Surely it couldn’t hurt just this once”, they say. “Even Jesus drank alcohol”, they say. “You need to let your hair down a little bit once in a while; it’s just not healthy to be so healthy”, they say. We kid ourselves with these little lies that seem to make sense, that seem so reasonable, and then someone comes who has been making the right little decisions for a long time, and we call them “talented”, we say they were “lucky”, it was “in their blood”, or maybe we outright accuse them of lying. Expletives cannot describe how angry that makes me – so angry that I can’t even get angry at it…because arguing with people who refuse to see sense only makes you stupider.

Anyway, back to the calendar, it’s basically a list of 18 months of days (540 days in total), dating from when I started Cantonese. Every day has a space for me to evaluate my SRSing, listening and reading, respectively. My task is merely to honestly evaluate and record whether or not I did my SRS reps, added SRS items, read some Cantonese/Chinese material and listened to Cantonese for the greater part of my waking (and maybe even sleeping) hours. X is “did nothing”, circle is “did it fully” and triangle is “half-done”. Doing SRS reps and additions takes 90 minutes or so, listening counts as “full” when it amounts to 10-12 waking hours or more, reading is 60-90 minutes. Listening can overlap with everything else, but for my purposes I consider SRSing and reading to be separate, if related.

I’m noticing that whether or not I do/live/play Cantonese has nothing to do with how busy I actually am, and far more to do with how organized I am that day. In fact, on my “perfect” Cantonese days (all circles), I have been berry, berry busy with other commitments and projects. Also, keeping Cantonese on while I sleep really helps. For one thing, it ensures that there’s no “morning warm-up”, whereby I forget to start doing my Cantonese immersion until, like, midday. It also gets me listening during my half-awake states (like just before falling asleep and just before waking up).

Victory Calendar

The last day on the calendar is fluency. Giving my fluency a date really makes a difference; it brings it from the realm of dream to the level of an actual calendar event. Maybe you can try making your own Victory Calendar 🙂 .

Indeed, one thing that drove me to go all the way with Japanese was that I had to be ready to go to a technical career fair at the 18-month mark, where I would have job interviews in Japanese. Money had been paid, air tickets bought and a hotel room reserved, months in advance. Cash and face were on the line. Through the Victory Calendar, I am trying to bring some of that “encouragement”, and concreteness, to my Cantonese process.

That was seriously me keeping it short.

[Edit: here’s a copy of my victory calendar].

  77 comments for “Chinese Project Notes 10: Big Developments (Anki, Text-To-Speech, Cantonese, Victory Calendar)

  1. Chiro-kun
    May 4, 2008 at 13:21

    Glad to see that your Cantonese studies are going well!

    Khatz, this link was posted in your blog long back (don’t remember which page):

    Some of it is similar to what you’re doing at the moment (audio+text). I’ve been wondering as to exactly how effective parallel texts are (Wan Zafran uses them, only without audio).

    What intrigued me the most was step 3:
    “3. you look at the translation and listen to the text at the same time, from the beginning to the end of a story, usually three times is enough to understand almost everything”

    I’ve been doing this kinda unconciously for a while (with anime and English subs) and it works really great and actually helps words stick! But if the transcript is available, parotting becomes easier too (ripping dialogue word-by-word sucks, and is really mistake-prone).

    I’m curious as to how effective this ‘method’ is. The closest substitute I can find at the moment are most episodes of One Piece (anime dialogue=manga dialogue). Then there are the subs which play the role of parallel texts.


  2. Jimmy
    May 4, 2008 at 15:13

    Wow…I had no idea that you actually made your travel arrangements and everything before you were fluent in Japanese. I always assumed you learned Japanese in a non-pressure environment (at least no pressure to learn Japanese), but even with that deadline hanging over your head, the fact that you reached your goal blows me away. 深く感動しました。Your ceaseless energy and positive attitude continue to amaze me. I suppose I could go the negative route and be intimidated by that…but I won’t. I’ll take it as a sign that I can do it, too, and I’ll work harder. Because I don’t want to use English anymore than I have to, I’m not going to leave any more comments for a while, but please know that all my encouragement goes with you on your Cantonese endeavors; I look forward to seeing you reach your goal. Thanks for all the help, and I

  3. Jimmy
    May 4, 2008 at 15:13

    ‘m sorry for the gushy comment.

  4. nacest
    May 4, 2008 at 17:04

    My kingdom for linux-compatible japanese tts software!
    (though this method sounds even more time-consuming than the kana to kanji review method… still, it sounds fun)

  5. khatzumoto
    May 4, 2008 at 17:46

    >still, it sounds fun)
    It certainly FEELS like it takes less time. For at least two reasons:
    1) My audio–>text or reading–>text cards are much smaller (shorter) than my other cards — generally 5-10 characters long. Too long and it’s too boring/difficult.
    2) It’s a lot of fun, really engaging.

  6. May 4, 2008 at 19:56

    Really encouraging Khatz, especially the calender thing is a good motivator I think. I’ll make one for myself aswell :).

  7. ghinzdra
    May 4, 2008 at 20:08

    Great idea to go for Anki !
    it seems you did it for the audio mod but maybe by now it occured to you the true asset of Anki lies in his card fact conception… It should be really helpful to you as in one of your previous post you acknowledged you could n t change your 30000 something cards to your new method ( kana to kanji) …
    good news owing to this conception card fact you cand do in the blink of an eye with anki… you just have to import your cards and then ask to anki to create both a recognition card ( from kanji to kana : your former method) and a production card (from kana to kanji : your current method) …. and it just the most obvious upshot of this brand new conception . As far as I m concerned I have gone further than that with several fields to really use at its maximum effencicy a single fact . For some facts if your fill out cleverly your fields you can create a vocabulary card , a production card , a close deletion card ,etc… from a single fact . I think Anki is making a impressive breakthrough for the SRS through this conception .

    On a side note I was looking at your book posts and I noticed the post about Understanding Basic Japanese Grammar by Koichi NISHIGUCHI [基礎日本語文法教本 西口光一(著)]. With friends of reviewing the kanji I put up a google group for mining sentences ( we called it AJATT sentences I hope you don t mind the homage) … We ve been mining Kanji odyssey 2001 which provides really interesting and very progressive sentences for 1110 kanji (as there are usually 3 sentences for each kanji it makes a lot of typing … but it s worth to do it : through this collaboration of 30 peoples it was really quick and this book cover 90% of the usual reading.) we ve completed the book within a month. Right now we re busy with checking and some japanese teachers I know have agreed on recording the sentences….and within the month we will tackle with another book maybe all about the particle or how to make difference between the japanese particle or another book…..So it really works great.
    But if we can spare even more time it s better and right now I noticed that your girlfriend typed all this stuff in her srs . Conditionned we can prove the ownership of the book would the both of you agree on sharing those cards ? I think our group can do with a book like this as I trust your judgement.

  8. May 4, 2008 at 21:01

    “Cantonese is even more “magical”, more BS-ed about, more Orientalized, more feared, more hyped than Japanese; this, I am sure, tickles my reverse-BS glands.”

    haha…that part cracked me up. and the “hearmegana.”

    don’t forget about the nihongo, though!

  9. ジェームズ
    May 4, 2008 at 23:37





  10. khatzumoto
    May 5, 2008 at 00:00




  11. ジェームズ
    May 5, 2008 at 03:25



  12. Yousef
    May 5, 2008 at 04:08

    Great post. I never thought of doing dictation, but I’ll try from now!

    I have nothing intelligent to say…but I just remembered a funny scene from the manga Nodame Cantabile. When Nodame first arrived in France, she had a lot of trouble learning French the traditional way. She finally learns the language by watching and mimicking over and over again her favorite anime dubbed in French. Someone in mangaland is getting AJATT (or should it be AFATT) ideas? 😉

  13. Mark
    May 5, 2008 at 05:51

    Excellent post! I love a good update…

    Good to hear about your progress with Cantonese – wow, how many tones does it have again?! And the Victory Calendar’s a great idea.

    Anyway, please keep it up.



  14. Godai
    May 5, 2008 at 10:41

    I really liked this post.

    Do you know/have any thoughts on ChinesePod?
    I’ve been curious about them for a while.

    Excellent job, man!
    Keep working!


  15. James
    May 5, 2008 at 12:18

    The audio files and some of the pictures aren’t working for me. T_T

    How did you make the calendar?

  16. Mark
    May 5, 2008 at 16:49

    Yep, couple of the pictures in this post didn’t work for me either. Haven’t tried the audio…


  17. Wan Zafran
    May 5, 2008 at 18:47


    As regards what you had said, I’d like to offer a correction. All of the parallel texts I make must be for novels/long stories that have accompanying audio. I do not peruse any text that come without audio. (I fear that my reading them in my own style may impress upon me substandard pronunciation and intonation.)

  18. Rob
    May 5, 2008 at 21:58

    Thanks for the SRS advice. About 2/3 of my cards have no audio and I was always wishing I had a native speaker around that I could record speaking my sentences. After checking out the Misaki Neospeech software, I decided to try it out and you’re right it is pretty sweet. Not only will it help with the SRS sentences, but it makes reading Japanese webpages more fun. If I come to a part that I can’t get through or want clarification, then I just click and Misaki will read it to me.


    Before going Japanese all the time, I dabbled with Mandarin and subscribed to ChinesePod. There is definitely a lot of information there, but I wouldn’t pay for their service like I did. In the beginning and intermediate lessons, they speak way too much English and even have non-native speakers doing the dialogues in some lessons. However, I do think it would be a great source for sentences to put into your SRS and you can do that for free.

  19. Forrest
    May 5, 2008 at 23:43

    Hmmm, I don’t see the Misaki voice on those pages… and it is the only one you didn’t link to.

    For Japanese female voice all I saw was Miyu and Kyoko.

  20. Kaede
    May 6, 2008 at 01:59

    First of all, this comment has pretty much nothing to do with this post (although the post is interesting, and those illustrations are lovely).

    Second of all, hi — found this site a few weeks ago through the forums. Been spending quite a bit of my free time at work since then reading through it, and like so many other people, I have to say thank you. I took 3 years of Japanese in college and loved it, although I didn’t really get as much out of it as I could or should have. Since graduating a year ago I’ve barely “studied” at all, which I’ve felt guilty about countless times. I’m never far away from Japanese — I listen to almost exclusively Japanese music, watch Japanese TV shows and plays, and try to plow my way through certain Japanese blogs, but…it’s never really been a concentrated effort, even though I love doing all of it. This almost ridiculously inspirational site has given me a plan, and made me feel like it’s one that I can actually stick with, and….that’s pretty awesome. So thanks. (I just wish I’d found this a year ago! All those months when I wasn’t working….and when I was feeling too guilty about not working to watch the Japanese stuff I really wanted to watch!)

    But ANYWAY. There is an actual point to this comment, and that’s to share something with your other not-in-Japan readers.

    Everyone knows about Amazon Japan (and their stupidly high shipping rates), and most people know about YesAsia (which I refuse to buy from anymore for petty reasons) and CDJapan (which doesn’t sell books as far as I know), but it seems like most people don’t know about HMV Japan — Their selection (CDs, DVDs, books/magazines) isn’t anything compared to Amazon’s, but for the things they do have their prices are usually around the same, and if you’re only buying a few things their shipping is far more reasonable. I’ve only used them a couple of times so far, but I’ve never had a problem. The site can be a little picky about search terms, so you might have to try a few things in both English and Japanese to find what you want, and there’s not much in the way of description so you really have to know what you’re looking for, but overall it’s not too much trouble, I think.

    Plus, they’re always having random specials — right now, they’re having a sale on a ton of DVDs, if you buy two or more. Included in the special are two of the shows that have been recommended on this site: <– Tiger & Dragon box set <– Trick Vol. 1 (I think?)

    So….that’s it. I hope it’s useful to someone 😀

  21. Rob
    May 6, 2008 at 02:54

    For anyone outside Japan and living in America or Canada that is looking for DVDs in Japanese, check out

    They are a company with offices in Canada and Washington (I think) and they import media from Japan. I got the 5 DVDs (Matrix, Apollo 13, Bourne Identity, etc) from them for under a hundred bucks including shipping. Most of their Japanese version DVDs run from $16-20 which isn’t far off from what you would pay for the regular version. Their customer service was excellent. I emailed their support to verify that all of the DVDs I was getting had the Japanese audio and subtitles and they wrote me back within an hour with the answer. I highly recommend them.

  22. nacest
    May 6, 2008 at 05:51

    Kaede, thanks for your recommendations.
    I’m curious about why you don’t like YesAsia. I was thinking of using it for the first time, but if you or someone else has had any bad experiences with their service I’d like to know beforehand.

  23. reineke
    May 6, 2008 at 07:23


    Are you sure about the new Cantonese-Mandarin pecking order? Is it still the old proven method? Are you watering down things a bit too much? I know you can do it, but is it wise? 🙂

  24. sven
    May 6, 2008 at 15:20

    hey Khatzo, this sounds very exciting. Ive got the tts for chinese and lily speaking! my question is to compose each sentence, you find the sentence/question, write it in tts, save it as mp3 then you import it into anki and then add the answer in anki. is that correct?

  25. sven
    May 6, 2008 at 15:22

    I am struggling with importing the audio recording mp3 files into anki. could you please give me the lowdown with the process you use, it would help me greatly. cheers

  26. sven
    May 6, 2008 at 15:38

    sorry, scrap the last two posts i was trying to import when i should have just been adding new cards and pressing the add audio button. this audio approach thing takes a bit longer but hopefully it will pay off in the long run

  27. sven
    May 6, 2008 at 16:29

    ive got the audio files in anki but they just don’t play…did you have this problem khatzo??

  28. David
    May 6, 2008 at 18:58

    @Sven – It’s a bug. They’re having some problems with the audio library; try using ogg vorbis instead of mp3.

  29. Rob
    May 6, 2008 at 21:22


    I don’t know if this will help, but when I first added the neospeech mp3 file to Anki, it sounded like Alvin and the Chipmunks. When I would play the file in media player, it sounded fine, but in Anki it would play at a fast speed. After some experimenting I found that if I had the audio recording setting in Neospeech to anything under 128 kbps, it wouldn’t play right in Anki. If you record at 128kbps or higher it should be fine.

  30. Jonathan
    May 7, 2008 at 04:07


    The speedup/slowdown problem with Anki occurs because it automatically plays every sound file at 44.10 kHz in order to be compatible with more sound cards. The bit rate of the file doesn’t actually have anything to do with it. Now, with Neospeech this doesn’t make much of a difference, since I think that 128kbps is pretty much the only encoding option that offers 44.10 kHz, but it’s useful to know if you want to put other soundfiles into Anki.

  31. rich_f
    May 7, 2008 at 05:17


    Misaki is not available for sale through the Nextup site. I talked to a Neospeech rep, and the base price he quoted me was $350 for the base corporate set. They don’t have a timetable yet on making a version available on the Nextup store. I don’t know if that means they don’t have any plans to do so at all, or if they just haven’t made any yet.

    It’s a shame, really, because I don’t have $350 to drop. I will say this, though, the Neospeech voices are by far the best ones I’ve heard. Misaki is great, Show is also very good. Miyu is a little robotic, but still very good.

    The other companies’ voices are all very nice… if you’re cutting a techno track and you want to put in a Japanese robot voice. 😀

    I’ll probably just spring for Miyu or Show for now. $35 I can handle. The tricky bit either way is going to be implementation. My main purpose isn’t so much to use it for dictation, but for listening to the sentences in my Anki deck with proper pitch accents. They never bothered to even mention that in college. 6 semesters, and not 30 seconds on pitch accents. Huh?!?

    Ah well. Thanks for the post as usual, Khatz.

  32. Rob
    May 7, 2008 at 05:30

    Thanks for clarifying that Jonathan. That’s good information to know.

  33. Madamada
    May 7, 2008 at 09:24

    I’ve been using Neosoft’s Japanese male voice for a while now and I’d just like to offer a caution.

    It lacks somewhat in terms of being able to select the correct reading in a given context. For example it reads 時 as じ when  とき is appropriate, 解く is read as とく when ほどく is appropriate. It is definately still worth using, but you have to be aware that you can’t depend on it to give you the correct reading and you may have to give it the kana spelling in some cases.

  34. rich_f
    May 7, 2008 at 12:00

    Hmm… yeah, I’ve seen that issue show up as well. But the thing is, はし and 橋 have two different pitch accents, even though both are はし in hiragana… I think there’s a dictionary editor in there somewhere, though. It may be time to hunt down the NHK pronunciation guide. I saw that at the local University library, and IIRC, it has the proper pronunciation for just about any Japanese word… I think. It’s been a few months since I was there.

    There’s an interesting discussion about similar issues here:

  35. rich_f
    May 7, 2008 at 12:21

    I don’t mean to overpost, but this looks hot to me. Too bad I can’t seem to find it anywhere:

    It’s a CD-ROM version of the NHK accent dictionary, with spoken word examples from NHK broadcasts of over 140,000 samples, and a dictionary of 66,000 words. @_@

    Sadly, it’s out of print. >_<

  36. ghinzdra
    May 7, 2008 at 17:31

    About this text to speech device I was so enthrilled by the outlook that I made research on this right on the spot .

    And Is there a best way to know how faithful to reality than asking the japanese themselves ? so without any warning I made listen news to several japanese teachers and collected their opinion.
    There were 4 teachers and 3 extracts : first one was Misaki reading a script of Fuji news Network , secound one the very newcaster of Fuji News Network , third one Yomira podcast .

    *everybody agreed on the enjoyable voice of Misaki .
    *Nobody agreed on the speed :
    the 1st one thought 0 was really too fast -7 was a radio speed , -6-4 a tv news speed , -4-2 normal speed
    the 2nd one wasn’t bothered at all by the speed I put it at 0
    3rd one frowned from the beginning . She sensed something unnatural
    4th thought it was a bit quick . -4 for her was a natural speed .
    * the hatsuon was perfect for everyone except the third .
    * when it comes to a direct comparaison
    the 1st one prefered Yomira over the other one but she still prefered Misaki over the real newscaster of Fuji News Network !
    the secund one was bothered by the lack of space but she thought it was due to needs of TV news . So she singled out Misaki .
    Only the third one really sensed Misaki as unnatural . But maybe she figured out what I was trying to do as she’s one of the teachers I asked to record the sentences for the KO2001 audio project . In the other hand she seems very sensitive to this particular problem as on the very beginning of the project she offered me to record also the vocabulary besides the sentence stating that the pronunciation really changes . After an explanation , she even pointed out part of the sentence where she thought it was unnatural .But the hatsuon of each word was normal according to her . She seems really reliable on this issue .
    the fourth one was also bothered the lack of space but until I told her she rejected the idea of Misaki as an artificial voice .

    So it seems like Misaki while not being perfect is really something reliable .
    . According to the people it ‘ll range from unnatural to normal but quick .
    (Once again ….:)) I just have to agree with Khatzu : you should look for a mix of natural voice and Misaki .
    In fact if you’re able to catch up with her 0 speed when you’re listening and if you’re able to speak like her you’ll be fluent for most of the japanese .
    As far as I’m concerned I planned to mix yomira podcast which is easy to download and to follow and Misaki reading the news of fuji news network (after having used a lot of software : goldwave ,getasfstream,etc… I eventually found how to get the streaming video file quite easily but it’s still longer and more troublesome than downloading the yomira podcast…. the strong point of FNN is that the script is totally faithful) .

  37. Kaede
    May 7, 2008 at 21:40


    I placed an order with them a little over a year ago for several things that hadn’t been released yet, and paid to have them shipped separately, because I wanted them all as soon as possible. The first thing was shipped out right after its release, no problem — but the next ones weren’t shipped out until after the very last was released, around a month later. And if I remember correctly, it was still some time after the last was released before they sent them.

    If I hadn’t paid for the separate shipping, and maybe if they hadn’t been things I wanted so badly, it wouldn’t have mattered, but….well, I said it was petty. 😀

  38. Rob
    May 8, 2008 at 00:47

    Thanks for that reseach ghinzdra. I wondered about the speed of Misaki when it’s on the default of 0. It seemed a bit fast to me too in some cases, but that’s cool because like you said keeping it at that speed can only help.

  39. nest0r
    May 8, 2008 at 03:40

    Hmm, I changed my mind, I no longer want to marry Fabrice for making the RTK site, I want to marry Khatzumoto for making this site.

  40. rich_f
    May 8, 2008 at 05:14

    One of the things I’ve noticed after messing around with the Miyu and Show voices, is that Show’s inflections are more animated– higher highs, lower lows, while Miyu’s are more understated. What I’m doing now is just copy/pasting sentences from my deck into TextAloud as I’m reviewing, then working on my pronunciation as I go along. I’ll usually listen to Show first at a slower speed, to get the basic idea, then I’ll see if I can keep up with Miyu at a slightly faster speed. Sort of using the parroting method to work on accent.

    Another thing I’ve noticed– there are still bugs. Sometimes 行きます is pronounced いきます, but sometimes it’s pronounced ゆきます. I know those are both valid readings of that kanji, and both verbs have their uses, but the way the program seems to decide is kind of random. One voice will say いきます, while the other will say ゆきます.

    I understand that the programmers are working on a new version of TextAloud, but I’m not sure if TextAloud is the problem, or if it’s the NeoSpeech pronunciation data. Not really sure who to report the bug to.

  41. Daniel S
    May 8, 2008 at 14:22

    So this may be… ok, is… a stupid question. But how did you make your victory calendar? Did you print it? Draw it on poster board? Where do you keep it?

    I love the idea.

  42. Rob
    May 9, 2008 at 23:59

    Is anyone else having the problem of not having much time for new sentence additions to the SRS using the hiragana to kanji / dictation methods? I started doing sentences in early March and since then have just under 200 sentences. The thing is, I do the reps everyday, but the time it takes to finish them usually leaves little if no time to find and add sentences. Should I perhaps create a schedule of say, addition only days and then have rep only days? Does anyone else do this?

    Khatz, with all of the improvements/changes to the actual SRS methods that have happened lately, perhaps a new post concerning sentence addition would be helpful?

  43. nest0r
    May 10, 2008 at 04:33

    I’ve been spending perhaps 1-2 hours on reps/additions together, doing TTS–>kanji. It was taking me forever, but then I realized that I’d unconsciously begun trying to memorize the sentences as a whole, so now I go much quicker. It also took some time getting used to switching back and forth between programs.

  44. nacest
    May 10, 2008 at 21:42

    Just to inform everyone, it is possible to run TextAloud and Misaki on Linux using wine. It’s not very easy to install at it has some glitches, but it’s usable for anki.

  45. khatzumoto
    May 12, 2008 at 10:31

    Yeah, first I need to add a “file output” feature to KhatzuMemo…

  46. khatzumoto
    May 12, 2008 at 10:32

    Never tried it…

  47. khatzumoto
    May 12, 2008 at 10:33

    Are the audio files and pictures working now?

  48. khatzumoto
    May 12, 2008 at 12:15

    >Should I perhaps create a schedule of say, addition only days and then have rep only days? Does anyone else do this?
    I recommend (this is what I do):
    1. Days of reps, and
    2. Days of reps + additions.

    So, always do reps. There’s no use filling a leaking bucket. Then, if you have time, do additions. In truth, you need to both add water AND plug holes in order to get your bucket full (get fluent), but always plug the leaks before adding water…I think.

    Most days I manage to do both reps and additions, but, if the sushi hits the fan, at least I’ve reviewed/plugged holes.

  49. dared
    May 13, 2008 at 02:35


    The strange readings seem to be caused by TextAloud, it looks like it doesn’t use the NeoSpeech pronounciation data at all. Unfortunately, I couldn’t find another program which can read highlighted japanese text at the use of a hotkey, so I use babylons voice feature to read me some words/sentences which TextAloud mispronounces.

  50. nest0r
    May 13, 2008 at 06:15

    Yea, I lied, OMG I totally lied, doing reps/additions every day is so hard. Think I’ll try doing additions every other day or something, good idea.

  51. Jim
    May 15, 2008 at 19:32

    Khatz and others: when doing TTS to a file, check to see what sampling rate your TTS voice supports. Mandarin Lily is 16khz. I found the best way was to do TTS output directly to a WAV file, then convert to an MP3 or OGG, maintaining the original sampling rate (16000) and mono. Otherwise you actually start to lose quality if you convert to 44.1 khz or even 32000! Try it and see the difference.

    The other issue is that Anki might play the 16khz file too fast. But don’t let that stop you, Anki will likely fix that in the future. For now, you can work around with a plugin: see the “” plugin. You just drop the “” file into your plugins folder, change the key line to something like the following:

    pygame.mixer.pre_init(16000, -16, 1, 1024*3)

    , and you’re done. Enjoy the perfect quality, small-size TTS files!

  52. nest0r
    May 16, 2008 at 01:21

    Whoa, thanks for the tips on the original sampling rate/.py file. Though now I have to separate the old decks I made from being played with that plugin, since it makes them slow motion.

  53. Ceryni
    May 27, 2008 at 12:34

    For learning chinese would Remembering The Kanji be useful? Or should one just trudge it out with some other book?

  54. khatzumoto
    May 29, 2008 at 16:56

    >Is it still the old proven method? Are you watering down things a bit too much? I know you can do it, but is it wise

    Short answer: wait till next year to find out
    Long answer: There is no “proven old method” per se…what I mean is, that old method was just one of a series of iterations . This is another, better suited to the situation at hand (Japanese can be accurately pronounced from kana, Cantonese does not use such a system), and I think better overall; it wasn’t change on a whim, but in response to a need.

    I certainly find this new iteration more engaging, and challenging in a good way, I mean, YOU try producing kanji from memory given only an audio cue 😀 . Haha, in fact, the best thing is to try it for yourself and see your results. I’m very pleased with mine, and of course I wouldn’t post it here if I hadn’t subjected it to some thought and experimentation and seen the results.

    Again, remember the situation:
    I read Japanese.
    I read Mandarin.
    Standard Written Mandarin/SWM and Standard Written Cantonese/SWC (yay! more TLAs!) are visually identical.

    I’m mapping three different languages to the same single kanji.
    All of these languages sound completely different but two of them are visually identical in their standard written forms. So there was significant internal pressure/bias/habit in and on me to vocalize SWC as SWM. Also, without a usable phonetic representation system, I felt like there was no good way for me to reliably *produce* Cantonese sounds *before*/without hearing them. In that sense TTS hits several clay pigeons with one shell —
    (a) I practice writing characters from memory and with only a sound cue
    (b) I make stronger character-to-sound relationships (earlier) than ever before, especially character-compound-to-sound
    (2) I overcome the bias to vocalize written Chinese in Mandarin;
    (3) I become familiar with the sounds of Cantonese i.e. it’s better for my pronunciation, since I’m always having a “native-level-user” “read” to me and of course
    (4) I naturally develop the ability to vocalize written Chinese in Cantonese (I imagine this is the part that worries you most: “if it’s being read to you, will you be able to read it yourself?”), the answer is: yes, after a few reps. Also,
    (5) I prevent over-dependence on visuals for comprehending Cantonese, a roundabout way of saying “I improve my listening comprehension and writing comprehension at the same time”. Working without sound tends to strengthen the writing without directly strengthening the listening…my ability to read Chinese was almost embarrassingly better than my ability to comprehend it spoken…of course more listening can help fix this, but, why not hit 5 clay pigeons with one shell? Much better than “visually learning” the text of a word, and then almost separately “aurally learning” its pronunciation.

    You know me, reineke, it’s Khatzumoto! Ah paid ja befo’! Sign the treaty!

    Yeah, but basically…wait around or try it.

  55. Ceryni
    May 31, 2008 at 14:20

    oh, just found the Rick Harbaugh book. Chinese Characters: a Genealogy and Dictionary. Would you think it best to go through it straight through, or is there some kind of order for easier learning?

  56. Mallory
    June 6, 2008 at 23:57

    Congrats! Welcome back!

  57. June 17, 2008 at 22:05

    Anyone know of some good collaborative learning chinese sites? is there anywhere where people are trading anki decks and recommending videos n stuff? sounds like japanese is all cool and popular, but i haven’t found any chinese-learning communities online yet.

    i’d like to work on building a nice deck of chengyu (idioms) and regular sentences, etc. ideas?

  58. AwkwardMap
    July 4, 2008 at 06:52

    I’ve been doing this (kana to kanji) for the past three weeks and it’s just a lot more fun. I’m playing a game here where I have to place the kanji in the correct spots all while still producing the kanji from memory (and not just repeatedly looking at the sentence as I write with the straight sentence method). I’ve not gotten TTS software yet (no money for it!), but that should change in a few weeks and then we’ll see.

    The image hack is also, yeah, the best way to learn nouns and such that I can just about almost get, but not quite just by the definition. Words like 衛星 and 頭蓋 are really easy to grasp with an image staring you in the face.

    So, yes, I think it very much so is worth doing with Japanese as well. I convert my failed cards into these along with the long ones when I break them up.

  59. July 7, 2008 at 12:43

    Khatzu…I love you =]
    I’m sure this is gonna work…thanks so much for making such a place…
    Rosetta Stone got nothin on you =]

  60. andy
    July 18, 2008 at 03:58

    I’m using your method for French. As to the SRS length, how long is 10 Chinese characters? Does that mean that I should limit myself to 10 French words or 10 letters? Thanks.

  61. August 26, 2008 at 03:19

    Hi! Wonderful site. I’m a Mandarin-learner, and I’m trying to get from “high intermediate” to “fluent” which is tough to do. Most textbooks/classes are too easy, and it’s really just a matter of exposure at this point. Your method is the best to getting there, I think, and I’m going to start implementing it.

    I also love your victory calendar. I don’t know if anyone else has asked you this, but would you ever consider posting it for download here? I started making my own, but adding all the days and such for 180 weeks looked daunting. (I’m due for a contribution to your tip jar, but I’d be happy to throw in a little extra for the doc, too.)

  62. ~
    August 27, 2008 at 01:11

    Wow, making the reservations and buying tickets before you even learned Japanese… that takes guts. o__o I think I’ll try that Victory-calendar thing. But what’s wrong with the BoPoMoFo? ;-;

  63. Markk
    December 11, 2008 at 12:08

    Heyy thar =D

    I’ve stumbled across your site and I’ve been applying it to my learning Cantonese :] I never really liked learning it much, cause all the stuff was reaaal boring: “你好嗎” “我好好。你呢?” 你呢” Stuff like that aha. I like your concept of “learn stuff you enjoy” because, in the end, im probably gonna be talking more about stuff I want to talk about than “Good day to you Mr. Wong. It is very sunny today. And how is your wife?” yaddayaddayadda. But what I really want to know is how long did it take you before you became really proficient at hearing and distinguishing/ speaking the tones? I’m having reaaal problems with the low-falling tone…
    Thanks :]

  64. Nostrum
    March 13, 2009 at 12:13

    Hey, I am using this victory calendar idea for Japanese, and I was wondering if anyone could tell me what appropriate headings for each part of it would be in Japanese… I don’t trust myself with something I’ll be looking at so often :P. The current headings in english are:
    Victory Calendar
    Date / Input / Countdown / Notes
    Review / Reading / Media / Add / Other [I have a lot of columns yes]
    and inside the chart I have “break” on certain activities for certain days.

    I was considering:
    date: 日付
    input: インプット活動
    notes: 注釈
    break: 休日

    countdown: not sure what word to use
    reading: not sure what form to put it in, if I want it to be a noun.
    and the other words as well.
    media, add [cards], other, need help for those too.

    any help in translating those words for my chart would be much appreciated.

  65. David
    April 20, 2009 at 13:35

    Hey Khatz, I wanted to let you how much of an inspiration you have been to my Japanese studies and methods. But there was actually one question regarding this post that I wanted to ask. Where did you find that Japanese voice for the TTS that you gave an example for? It sounds phenomenal but I can’t seem to find it on the link to the website that you provided. All I’ve found is Kyoko and I don’t think it’s the same and the one you showed.


  66. WangSen
    December 4, 2009 at 18:11

    Hey Chinese learners…I have a simple question for you guys (and gals).
    I have been learning my hanzi with Anki. I have done pretty good sticking to it with the occasional life interruptions, but I have tried to keep those to a minimum. Once I stopped for a month, so I reset Anki and started from the beginning. Now I have 3000 hanzi down well. My question is rather simple. Instead of continuing to 4000 hanzi, does anyone see something wrong with starting to do sentences and just adding new hanzi as I come across them from this point on?
    I think it actually maybe beneficial since I would be sure that hanzi I would learn this way would be learned in context. Any thoughts?
    Thanks all…and thanks to Katz..

  67. Mike
    July 26, 2010 at 13:15

    I have some experiences with text to speech software programs, I recommend Panopreter Plus
    ( to Windows 7 or vista users, The software reads text and converts the text to mp3 and wav files with the voice of Microsoft Anna. Best of all, it’s easy to use and useful to language learners.

  68. Jonathan
    July 26, 2010 at 18:01

    Your blog is awesome and of course you are good looking and I want to have your babies.

    At any rate, I think you and I are cut from the same cloth. I’m currently living in HK and am learning Canto since “you will never learn it; too hard”, blah, blah, blah.

    If there are any Canto resources that you can share, I’d be greatful.

    Keep up the awesome work.

  69. Puchatek
    September 5, 2011 at 11:27

    I know this thread has been dead for a while, but I found a really neat (and probably not all that new) TTS solution for Anki and felt it’s worth sharing should anyone stumble upon this page again (like I did not that long ago), There’s an Anki plugin called Google TTS or similar that lets you very easily turn Japanese (and other) sentences to mp3s. It requires connection to a network, and when it screws up pronunciation (i.e. 中 as なか and not じゅう when it really should be じゅう) you can’t do anything about it, but it’s free, straightforward and hassle free.

  70. Lane
    February 3, 2013 at 10:32

    I didn’t know about ‘victory calendars’, but I essentially have one. I’ve got 3 years left in school and I’ve got a trip to Japan planned every December. I’ve got a spreadsheet with a line for each week and I have different columns for different weekly sentence addition rates (80,100,120,etc). I tend to add 120 a week. I have a creeping section where I track how many I’ve done with a black background and I have my target for this week in a red background.

    It’s seriously motivating to see where I *could* be at x, y, and z dates in the future (with highlighting on my ‘test periods’ when I’ll be in Japan at the end of each year). If I’ve done 80-100, then “all I have to do” for that extra benefit down the line is a few dozen more sentences this week. It’s a tradeoff I’m always willing to make with myself… especially considering where you described your skills after 7500 sentences and where tkyosam described his after 8700.

    I mean getting jobs that rely on your language ability is a legit measuring stick in my book and in your case as well as in tkyosam’s, you landed jobs well before the 10k mark. So here’s shooting for 10k and we’ll see what shakes out. (I’m shooting for 800 this week and already seeing improvement)

    Cool advice. Cool blog. Thanks for sharing the wealth.

  71. Raphael
    October 24, 2015 at 17:03

    Does anyone know about a good Korean TTS? I’ve been using imtranslator, but find it to be too robotic.

Leave a Reply

Your email address will not be published. Required fields are marked *