Today Zteer (the creator of bkh01 and advanced pitch editor) released one more UTAU engine. This one is based on EFB-GW, but differently from it, it WORKS (at least, worked for me, I'll develop this part a bit below there).
It can be downloaded on this site (along with some other projects made by Zteer): http://z-server.game.coocan.jp/utau/utautop.html
So, time for the review:
I think this engine sounds quite unique so far, and renders relatively fast for me. It seems that it doesn't make any special frequency file, simply using the already rendered .frq files from the VB.
Although not sounding very close to TIPS, it behaves a bit similar in the matter of range. In the cases I tested, it made the voice reach C1 with no problems (it was a male voice, recorded in C3); and makes higher notes sound a bit breathy. This breathiness limits the upper range, though. The same voicebank I tested only sounded well up to C5, beyond that, he sounded too breathy.
tn_fnds doesn't accept ANY flags, not even breath control. But accepts acpt use. Formant filter can't be turned off either. This is a bit of a drawback, cause you can only work with the raw voice.
But, for the other side, it makes longer notes in a way that none of the previous resamplers did. It loops a period, instead of literally stretching the sample. The good point is that this reduces the metallicness on the voice A LOT, BUT, voicebanks with short samples will sound bad with it (Defoko is doomed on this resampler).
I found a small bug, but nothing that would make it impossible to work, and is related to my English VB only. But so far, it works pretty good.
Now, the point of it working... A friend of mine downloaded it, but couldn't use. I don't know exactly the reason. But I THINK it was because he doesn't have Japanese locale, so I'd like to warn people that there is some chance it will not work. And also, probably like the previous resamplers, it may work better with some voices, and others, not-so-well.
It also requires downloading a bunch of fftw dlls, the link is included on the topic of the engine itself, on the download link provided.
tn_fnds works exclusively with mono samples. If a sample is stereo, it will give a error message, and produce no sound. Also, as default for UTAU, it is also recommended to use samples with 16 bit for bitrate, and 44100 Hz for samplerate.
Here is the test I made with it:
[soundcloud]http://soundcloud.com/tady159/migihidari-dashi[/soundcloud]
It can be downloaded on this site (along with some other projects made by Zteer): http://z-server.game.coocan.jp/utau/utautop.html
So, time for the review:
I think this engine sounds quite unique so far, and renders relatively fast for me. It seems that it doesn't make any special frequency file, simply using the already rendered .frq files from the VB.
Although not sounding very close to TIPS, it behaves a bit similar in the matter of range. In the cases I tested, it made the voice reach C1 with no problems (it was a male voice, recorded in C3); and makes higher notes sound a bit breathy. This breathiness limits the upper range, though. The same voicebank I tested only sounded well up to C5, beyond that, he sounded too breathy.
tn_fnds doesn't accept ANY flags, not even breath control. But accepts acpt use. Formant filter can't be turned off either. This is a bit of a drawback, cause you can only work with the raw voice.
But, for the other side, it makes longer notes in a way that none of the previous resamplers did. It loops a period, instead of literally stretching the sample. The good point is that this reduces the metallicness on the voice A LOT, BUT, voicebanks with short samples will sound bad with it (Defoko is doomed on this resampler).
I found a small bug, but nothing that would make it impossible to work, and is related to my English VB only. But so far, it works pretty good.
Now, the point of it working... A friend of mine downloaded it, but couldn't use. I don't know exactly the reason. But I THINK it was because he doesn't have Japanese locale, so I'd like to warn people that there is some chance it will not work. And also, probably like the previous resamplers, it may work better with some voices, and others, not-so-well.
It also requires downloading a bunch of fftw dlls, the link is included on the topic of the engine itself, on the download link provided.
tn_fnds works exclusively with mono samples. If a sample is stereo, it will give a error message, and produce no sound. Also, as default for UTAU, it is also recommended to use samples with 16 bit for bitrate, and 44100 Hz for samplerate.
Here is the test I made with it:
[soundcloud]http://soundcloud.com/tady159/migihidari-dashi[/soundcloud]