As khatzu suggested I think it's of great help of using the TTS technology with the SRS . I dambled a bit with it for a while : gave up , resumed , gave up , resumed....
before I go forward I must point out that I'm using Text aloud as my main TTS software and Misaki as japanese voice.

Now as I realized my listening comprehension is weak (especially if I draw a parallel with the reading comprehension I reached with Anki-KO2001-KM 2 Kyuu) I m serious about it . I'd like to create an audio file for every and each sentence .... After a quite painful process with my database and a bunc of software like calc , antrenamer, a split software , etc.... I eventually came up with thousands of text file, perfectely ordered , one by sentence . I intended to use the batch function so that I might avoid to take each sentence and put in my TTS software ,correct some mistake , change the name, record it , etc.... Unfortunetaly when I used the batch function
it turned out that it was totally ineffective with the misaki voice .... I mean I did get something but it was no human language , not any that I know at least.... more something like the eerie voice of a baby and a woman stuttering ...
so to sum it up :
-it works perfectly If I copy the japanese text with the misaki voice in the window and read it or record it (and of course it works too with an american text)
-it works perfeclty If I batch american files ... (with the peter voice for instance)
BUT it fails if I use the batch with japanese file : I tried to change name of the file , type of file , compression , .... Nothing works.

I wondered if anyone else encoutered this kind of problem ?
Otherwise do you use an another software with better results : I gave a shot at speak text , text aloud , free natural reader , text sound , 1st read it aloud....
text sound doesn't seem to allow to include new voice (so no miyaki)
1st read it aloud has no batch function

when it comes to the 3 others while they have both a batch fonction and a large array of voice (including the misaki ) If i use the misaki voice it just crashes or give no result . As I'm using trial version it's possible that it's accountable for those troubles I can't be sure .

I did the same as you and had the same problem. The textfiles have to use one certain encoding (which I forgot). Try opening one file with the text editor and save it once in Unicode, Unicode Big Endian and UTF-8. One of those should work. If you found the right encoding there are some tools to batch convert all your files to the correct encoding.

If I were you, I wouldn't be THAT dependant on Misaki. The software IS amazing and can pronounce most sentences brilliantly, but even then it's pretty far from native Japanese and sometimes it's completely wrong. If you do it in a batch function, odds are you're going to get lots of sentences where Misaki puts spaces in the wrong places etc and if you use them to study, you'll learn wrong.

Best idea IMO is to use it sparringly, and only manually so you KNOW she has identified the words correctly.

I'm using it manually, so I'm confident that I know when Misaki's messing up. I don't want to lead myself down a wrong turn, but it's just so nice to have audio in Anki, mixing things up a bit with the kanji.

Hopefully the AJATT thing should smooth out any wrinkles that get through.

Just hire some poor high school student to read the sentences in for you. Pay them 100$. I'm sure you could fit in at least 1000 sentences in one session.

stehr Wrote:Just hire some poor high school student to read the sentences in for you. Pay them 100$. I'm sure you could fit in at least 1000 sentences in one session.

Genius. I can see it now "My TTS is a herd of teenage Japanese slaves I keep in the basement..."

You'd quickly become an expert at such sentences as "I'm hungry", and "Stop it, it hurts!"

Sounds like a good job for the mechanical turk. https://www.mturk.com/mturk/welcome

thanks a lot sequa!
it works just fine .It's unicode by the way (UTF-8 will only give those britney spears lyrics I've been bitching about) . Too bad I had to go through another round of chores.... (most of the text replacer don't deal with kanji .... They mess it up in a way or in another so I had to take it all way back to my excel file and register it this time as unicode)

On a sidenote I do know the backside of relying only on misaki. That's why I hear a lot of "natural" japanese (news , anime, etc...) . And I also intend to take it reverse way : I have cds of another manuel I don't want to typewrite so I ll find a speech to text software .

Regarding using Mechanical Turk to get human recordings for pay:

I've successfully used MTurk to get a recording of a single Chinese text of several sentences; it cost me $1 and took 12h turnaround, though for bulk you'd probably be able to pay less. If anyone wants more info and my HIT template, PM me and I'll post it.

Japanese-speaking MTurk workers might be a little harder to come by (as there are fewer Japanese immigrants in English-speaking countries), but I'm generally hopeful.

How to use it for many single SRS items: So I asked the worker to upload the recording on rapidshare and submit the link, which worked fine for a single text, but it's too cumbersome (= expensive) if you're going to get individual recordings for a lot of sentences / words. So either you'd need some kind of recording applet on the HIT page (quick googling doesn't give me much hope to set this up in a less than a day without paying for it), or you could just combine bulks of ~100 sentences each into single large HITs, asking the worker to record a lot of files and ZIP them (or to make a long recording with 1.5-second pauses between items so you can split them with an audio editor).

Then just figure out how to get a bulk of numbered audio files into Anki. Play with the import function and perhaps use a script, I guess.

Ugh, so much work. I'll definitely go finish the KO2001 Anki deck with audio before I attempt anything crazy like this. ^_^