Comparison of Female English speaking voices for Synthesizer-V
HTML-код
- Опубликовано: 28 окт 2023
- This is a quick demo of the four female voices that Dreamtonics considers English speaking. The voices are Solaria (Eclipsed Sounds), Sheena (Dreamtonics), Weina (Dreamtonics) and Natalie (Dreamtonics). I've put them in order of timbre, starting with the most intense (Solaria), down to Natalie (most relaxed).
The point of this demo is just to give you an idea of the differences between these four voices, so you can decide for yourself where the crossovers are and which would be most suitable for what you're doing.
The track is an excerpt from a song I'm writing called "Dreaming of Tomorrow". It's actually being written for Asterian, but it works for these ladies too. 
Видеоклипы
![[SynthV Cover] The Phantom of the Opera [SOLARIA & ASTERIAN]](http://i.ytimg.com/vi/ZhGwWzw27kc/mqdefault.jpg)
Great video! Having a comparison like this is very useful to help me decide who I should buy next between Weina and Natalie. Just wanted to add some 3rd party female English AI vocals for SynthV that I didn't see mentioned here and my personal thoughts on them just in case it's useful for anyone else:
ANRI (made by Audiologie): To put it bluntly, Anri in her current state is not something I would recommend buying. While part of it is due to her being an older vocal, other vocals released around the same time as her still hold up incredibly today (Solaria was released only a month later!). She's very muffled, her vocal modes don't do much to change her tone of voice, and it takes a fair amount of pitch/tension/etc editing to make her sound good, much less realistic. However, she is receiving a new, updated voice library called ANRI Arcane soon that does appear to solve many of those issues, so if you're looking for a younger woman with a light tone of voice, that new update could be for you.
Eleanor Forte AI (made by Volor): Eleanor Forte started as a non-AI vocal, but her AI update allows her more realism while still keeping some of her previous charm. She's got a slightly accented voice, but it feels more like an accent a native English speaker could have. She's not necessarily as realistic as a vocal like Solaria, but she seems to work extremely well as background vocals for electronic pieces. She's got more of a calm, subdued voice, but she still has the potential for power.
Tsurumaki Maki AI (English) (made by AH-Software): Tsurumaki Maki got released at a weird time when crosslingual synthesis wasn't an option yet, so she ended up getting separate English and Japanese AI libraries.... and then crosslingual synthesis happened, making them obsolete and a waste of money. Whoops! Her English library is still noticeably accented and isn't intended to be realistic, instead aiming to be sort of a character voice. Her tone is very nasal, so if that is something you dislike, you're probably better off not buying her. If you're looking for a teenage-sounding singer, though, she's not the worst option you've got.
I could talk pretty in-depth about the non-native, non-English AI bank's ability to sing in English tbh, but this comment has gotten long enough so I'll stop typing. I hope something I say in this comment can help someone decide how to best spend their money :)
[disclaimer, since I'm pretty sure I'm not totally the target audience for this video. I'm a hobbyist and a huge fan of vocal synthesis technology since the Vocaloid 4 days so what I personally value in a vocal is probably different than what someone attempting to make professional-quality music might value. I personally enjoy a lot of the weird, sorta accented results you can get from crosslingual synthesis in SynthV, but I acknowledge that it won't vibe with everyone.]
I must admit, I'm pretty sensitive to accents. I don't think they're that important for lead vocals, because they should have lots of personality, but backing vocals need to blend seamlessly. If you hear an accent in backing vocals, they're not really doing their job properly. This might be specifically an American/English problem, I'm not really sure. Or maybe its just a me problem. I feel that in songs that originate from US/Canada/UK, we expect backing vocals to not have an accent, unless they match that of the lead vocal. I may be wrong on that, but it has been my observation in the music I listen to. Anyway, the point of this "shootout" was really just a quick generic side-by-side, since they seem quite sparse on the web. When I was thinking about getting Weina (which I only did recently), it was hard to make a judgement call on the current state of development of that voicebank, even though it had been out longer than most. So I know these things are pretty useful. Anyway, thanks for your contribution, which is welcome.
Great comparison, and a very nice song! If you're at all interested, there are actually three other female English AI voices available for SynthV that are available if you go onto their companies's respective websites. I'll provide a brief description of each vocal, if any of them sound interesting to you! They're all more so targeted at the "anime" side of the program's audience, but are definitely great voices in their own right.
1. Anri Arcane from Audiologie. A great youthful voice with a lot of personality, but can sometimes have fairly plain phrasing. Pretty diverse in terms of tone, although I find most of her vocal modes to be superfluous. An upgraded version of the original Anri vocal, which is no longer available.
2. Eleanor Forte AI from Animen. A lovely soft and mature vocalist with very delicate pronunciation. In older versions she sounded a bit unnaturally "thick", but she's improved with updates. Unfortunately, not a very tonally diverse voice, and can be somewhat difficult to use. Also has a lite Standard voice, but it sounds completely different and can't be used in anything monetized.
3. Tsurumaki Maki AI English from AH Software. A native Japanese speaker, but the voicebank itself is English. She was created before crosslingual synthesis was a feature, and therefore had separate Japanese and English voices in both AI and Standard form. Has a cute, somewhat raspy tone. However, she can sound almost as mature as Natalie with her Adult vocal mode and some edits. Has a stronger non-native accent than any other English vocal, so be aware.
Bonus: Stardust from Quadimension. Not *officially* listed as an English vocal, but has English recordings that make her sound much more natural in English than other Chinese vocals. My absolute favourite vocal for the program, has an even more diverse tonal range than Solaria and can be used in just about any genre. Has a very distinct husky tone. I wouldn't recommend trying to purchase her unless you're *absolutely certain* that you want to, because she's even more expensive than Weina and is extremely difficult to purchase from outside of China, usually requiring a proxy service that'll charge you like $30 to forward an email. Also has a very annoying and strict commercial license.
Thanks for adding your review of these extras. I've never considered them because of their vocal characteristics, but they do all deserve to be included. It's hard to find good demos of these voices doing original songs. I've heard a couple of Eleanor Forte AI that sound pretty good. I think the others just go too far on the anime side.
Thanks for making this video, my friend. I was able to find the perfect voice for my song.
Which one did you go for?
@@themidimachine I'm still considering it, but I think I'll go with Solaria or Sheena. By the way, I usually use Mai.
@@ciroto I imagine a bit of a conundrum for you. Solaria definitely better for English speaking... I imagine Sheena much better for Japanese since that is her native language. Tasty guitar playing on your channel btw.
@@themidimachine I see. I can only speak Japanese, so I couldn't understand native English. thank you for teaching me! And thank you for watching my video.😊
Nice comparison...very useful. I have solaria and weina but Natalie sounds like a nice voice also and perhaps more useful for my stuff than weina
I think for pure English speaking stuff, you can't beat the Solaria/Natalie combo for a really wide range of material. They are significantly different from each other. For now at least, they will remain my two go to's.
Ooh interesting comparison!!
The song is beautiful! I hope to hear the full version
Thank you @zihiko. I accidentally deleted my DAW file so I no longer have anything to work from, but before I did that, I did a version that's close to what would have been final... audio only at alonetone.com/richiebee/playlists/all-mine/dreaming-of-tomorrow, and the final uses a male voice (Hayden). Lyrics are also posted at this link.
@@themidimachine oh, too bad:( But thanks! Hayden sounds also sounds so realistic and good
Thanks for making this video! I think solaria sounds like the best singer, but to me Sheena is has the nicest sound in her voice . I have Solaria and without soft/breathy sliders higher up, it sounds so sharp and kind of like she's yelling, which reminds me a bit of the singer who sings the original "let it go" frozen song. Such voices can easily sound a bit " bitchy" , whereas I find Sheena sounds like a friendlier voice.
However, the Sheena accent is a bit disappointing though. I wonder if there will come some Ai that can alter or remove accents though. However, that would require extra hassle.
What I also like about solaria is that I find that I do not have to tune much, which is my preference. However, some say solaria still needs tuning. I'm wondering if you agree with that? To me it often already sounds natural out of the box. However, I',m not a native speaker and compared to others I may not listen as closely to voices in general , as to me a voice is just some instrument that I have at my disposal and listen to all instruments somewhat equally (some people are very voice focussed and listen to music because of a voice whereas for me instrumentation and composition are much more important).
I'm also disappointed by Sheena's accent. Natalie used to have a bit of one, and now I don't think she does, so it is possible that they could tweak some of it to reduce the effect. I think Asian voices though will always have issues with some English syllables that they don't have in their own language. I think Solaria sounds natural out of the box. I do very little tweaking within my own songs. A little emphasis or increase or decrease in intensity, but I don't do it at a macro level, and I think Solaria (and Natalie for that matter) do phrasing very naturally.
@@themidimachine Nathalie used to have an accent? But which accent did she have? I mean,everyone has an accent (americans/brittish people) ..just different ones.
@@PowerRedBullTypology It wasn't pronounced, but it was there... kind of made her sound a little um, Eastern European... it was only noticeable in some phrases, but when it was there it could be annoying. It's completely gone now. I never noticed it when I first got Natalie, but then working on a bunch of songs, you start to notice it, and then you can't unnotice it. I stopped using Natalie for that reason, but she's back in favour now that's gone.
@@themidimachine Great they are able do make such thing disappear then, apparently. If it was done with technology, you'd expect that newer voices do not have it at all, but like you said, maybe if letters are not there in ones natural language, it may be hard to replace wrong pronunciations?
Thanks - which one is best for a beginner like me? Solaria seems to be the most comprehensible, but is there any difference in how to use/tune the voices?
They have similar controls, and you'd approach using/tuning the voices in the same way. Some have more modes than others, but they all basically have controls that make the voice softer/breathier and harsher/tension. For English speaking female voices, I would say Solaria and Natalie are the best choices for clarity and diction. Tonally they are very different from each other, so I'd pick the one that best suits your style of writing. For pop, I'd say Solaria is the best. Male vocals have typically had a number of quality issues that female voices don't have. This is improving quickly, but I'd say the best male English speaking voices are Hayden and Saros - these are very similar in many ways - but the newly released Saros has the edge on both pitch and dynamic ranges. Overall, I think the female voices are easier to work with than the male voices.
Excellent demo, really helpful, thanks. I think Natalie would be the one for me. Is it easy to input syllables at different pitches? Eg, Diff-er-ent. Do you just write each syllable as you would a word?
It's very easy. It's part of a workflow. I can put together a quick video tomorrow night if that would help. The workflow I use is incredibly quick and natural.
@@themidimachine That would be really useful, thanks. I’ve subbed.
@@Tolbiny Hope this will explain it... ruclips.net/video/9zibPuKszpw/видео.htmlsi=BfqeyU3idpSnqS1V
I know that Mai is not on pair with those voices because it lacks various modes, but somehow i feel that she had almost no accent (definitely better than Sheena and Weina) even it's Japanese voice originally. However, i'm not native English speaker so maybe i'm wrong, don't know. Still, Mail can achieve very nice tone colors with a bit of play with parameters (her default voice is kind of funny to me).
I don't think that lack of vocal modes is really an issue, since you can do so much with Parameters, which are the same for all voices, though Mai never really goes away from soft I suppose. I should have included Mai in this test, and I think I only didn't because I didn't even think about her since she's not for sale on Dreamtonic's web site. I think her accent is similar to Sheena's (not surprising since they are both Japanese). I used Sheena as lead on a song recently, and because there weren't many syllables that had an accent, her voice works perfectly for the song. In fact, on songs where her accent isn't noticeable, I think she's my favourite female voice. I should try it with Mai. I don't like Mai's voice by default, but if you drag the gender slider to the right and up the tension, it's decent.
Yeah, first thing i do is to move gender to somewhere between 0.150 and 0.300, and adjust tone shift to match desired tone color. A agree about Sheena and her really nice voice - if you avoid specific phonemes - like "ch" in "feature" in this short demonstration. In your new song there's no "ch" so i didn't noticed any issue with accent, she fits great :) Unfortunately i'm completely amateur and playing with this only for fun, so i can't afford all those voices to play with, but from all videos i saw until now, those better voices are easier to use "out of the box", while Mai still need a lot more work to sound really good, and as you are moving away from defaults, voice is not so clear anymore.
@@vladimirzivkovic3393 I'm an amateur playing with it for fun and personal entertainment too. I spend all of my disposable income on music stuff... I'll never make even a fraction of it back. LOL. It's a good job they don't come out with new voices every month.
Thanks for the comparison. I’ve been trying to get a similar sound out of Solaria with not so good success. I’m still trying to figure out why.
Is it possible that you’d upload the svp file? If not, which vocal modes did you use for them?
No use of vocal modes - except Solaria, where I accidentally used a "Soft" mode of 10%. Also I tried to balance levels somewhat, using the Loudness setting - boosting or cutting voices by no more than 1.5dB. This was more for the harmony bit at the start of the video than anything else. The point was to use them at their default settings on loading. I did start a second test of trying to match the timbre, or more accurately, the tension to Solaria. I never finished that part of the video because I specifically wanted this up before the end of the weekend, but I have the recordings, and will hopefully have a chance to finish that off over the next couple of days. It's perhaps a bit unfair to compare English/US voices with Asian voices, since the formation of words, right from when you start learning to speak, is fundamentally different, so trying to grade them based on English, is unfair, but its still fact, that the two natively English speakers (Solaria and Natalie), have much more distinctive voices with more personality, when sung in English.
Nicely done comparison video, perhaps one with less effects next time.
Solaria: ruclips.net/video/pPSJAZvbhZ8/видео.htmlsi=JVnbH0wrgG0MTJko
I use Synth V and find it very useful. How’ver to call voices English is a misdescription. They are American with American pronunciations which toEnglish ears can grate.
No matter how much you think other nations butcher the language, it is still English. You can change the pronounciation of some words to make them English sounding, but you would certainly expect an American singer, for example, to use American-English.
@@themidimachine Sorry, but I am afraid that what you say is not quite right. I hold a qualification from Cambridge University to teach English to foreign students. While the UK has regional accents and dialects there is a standard form of English called “Received Pronunciation” sometimes called “The Kings English”. This is what we teach.
Not only is the American accent totally different; there are different pronunciations and often different spellings of the same word, as well as different construction of sentences.
To Western ears Chinese and Japanese sound the same. Are they?
Yours is a fantastic product but English is not American and I do recommend that you distinguish between the two.
@@neiljopson8138 The title of the video refers to English as a language, not an origin. Americans speak English. So do Canadians, Australians, New Zealanders and people from a whole ton of other countries.
@@themidimachine I have obviously failed to explain myself clearly. All I suggested was a singer or singers whose accent is not North American and Received Pronunciation is as neutral as one can get. End of!
@@neiljopson8138 No idea what you're trying to say here.