That's pretty common among TTS services and engines. Most people want a female voice for their personal assistant/ebook reader/whatever, so the male voices don't seem to get as much tuning. As someone who happens to prefer a male voice on my apps, it's a little frustrating.
I don't know how much is tuning and how much is personal preference but I plugged in some different texts and I'd probably default to Amy (English, British).
Do you have a reference to the restrictions of what is prodced by "say"? Just because a tool may be GPL does not mean that its product need be (c.f. gcc)
> F. Voices. Subject to the terms and conditions of this License, you may use the system voices included in the Apple Software (“System Voices”) (i) while running the Apple Software and (ii) to create your own original content and projects for your personal, non-commercial use. No other use of the System Voices is permitted by this License, including but not limited to the use, reproduction, display, performance, recording, publishing or redistribution of any of the System Voices in a profit, non-profit, public sharing or commercial context.
Which truly sucks, because they don't even give you the option to pay for such use.
I noticed the Brazilian Portuguese voice examples have URLs with the word "hola" (Spanish), and the text starts with Olá (more formal, used in Portugal, whereas Brazilians prefer the informal Oi).
Native Spanish speaker here. Is it just me or is the Spanish recording much worse than the English one? I don't speak any of the other languages to make a fair comparison.
I wonder why they pulled the Ivona text2speech android app. I'm still happily using it. It's quite comfortable to listen to articles in pocket while on the train (and to keep listening while changing trains).
Does anyone know if NPR recently used this (or something similar) recently, on a story? I recall listening to a story in my car (hardly paying attention) and the person narrating the story, sounded unusual. I thought it is probably computer generated but I couldn't tell for sure. I guess the general public better get ready to tell the difference?
You can read this as: We're pleased to announce our really cool TTS feature that took a lot of engineering know-how and effort ... but you'll have to click through, because we can't seem to get around the limitations of our CMS to embed audio content in a blog post.
I tried a few random sentences and some articles paragraphs with both Vitória and Ricardo (Brazilian Portuguese voices) and Ricardo did pretty well. I was impressed, really. Vitória on the other hand was not much better (as in "fluent", with rhythm and right intonation) than other available female voices out there for pt_BR.
If you have a podcast, at https://www.podigee.com we're about to launch a beta transcription service that is integrated into the podcast publishing workflow. Drop me a line if you want a beta invite for the feature.
Wish we had seen this a few days ago before dropping funds on a human to record for us. I played with several different voices and ran it on a text corpus that we gave to the human, and in some cases I would say this even sounds better.
Computer generated voices feel most robotic when their intonation of a word is abnormal or their pauses between words make the sentence feel choppy. The intonation and natural pauses between words is very good for all of the main voices.
The Japanese voice Mizuki was the most comical addition, since I can't think of a real situation where she would ever actually be used. Mizuki speaks Engarish (the Japanese version of English) beautifully, but any Japanese person who can understand Engarish will also understand English. Also, Mizuki doesn't add the correct vowel ending to all words, e.g. she correctly says "cheezu" for "cheese", but says "steku" instead of "steki" for "steak".
>any Japanese person who can understand Engarish will also understand English.
My impression is that it's actually quite common for Japanese people to understand Japanese accented English more easily than a native English speaker.
>she correctly says "cheezu" for "cheese", but says "steku" instead of "steki" for "steak".
"Suteeku" sounds like what a Japanese person who knows English well but has a strong accent might say. "Suteeki" is more of a corrupted loanword.
I would say it's true in general for us non native speakers to understand other non native speakers better than native speakers (unless their English is so garbled that becomes too hard...). But it's even more especially true for English pronounced by people born in our own country; that's the easiest by a far amount
> My impression is that it's actually quite common for Japanese people to understand Japanese accented English more easily than a native English speaker.
This is actually very true, made me think back to many situations in Japan where adding a strong Japanese accent to my English words made it comprehensible to the listener (just as adding a Japanese accent to the generated speech makes it nearly incomprehensible to a non-Japanese speaker).
One thing I wish more services like this offered is non-speech sounds. CereVoice, for example, lets you insert laughs, coughs, sighs, etc and it can really enhance the output in some cases. Google's WaveNet also manages to simulate the catching of one's breath during particularly long utterances, although I realize it uses a completely different technique (neural net vs. concatenative synthesis).
My biggest problem with CereVoice, though, has been its terrible web API. It doesn't support streaming output, so it renders the audio to an Amazon S3 bucket and then returns a URL, which is pretty inconvenient (and slow). You have to do the same for transcripts, too. So, if you want everything, you have to make 3 separate HTTP requests and parse 2 XML documents for one round of synthesis.
IBM Watson's TTS API gets it right, imo. Its streaming mode returns audio frames and transcripts over a WebSocket connection.
I was trying to figure out an affordable way to send "Read Later" articles in voice to mobile device, either as podcast or other format, to keep myself relevant while driving to/from office.
I realized this tool might not be cheap, since it may take the voice actor/actress 2 hours per day to produce my content (2-hour driving commuting per day for me). To get familiar local accent, it costs ~$36 in Australia, and maybe slightly cheaper for US accent. The value it brings me can hardly justify the cost.
Now, with Polly, things changed - it produces reasonable voice, and 2-hour content would only cost ~$0.3. I decided to launch my service as soon as Instapaper approves my API request.
Different optimizations. WaveNet (depending on hardware and configuration) takes hours to tens of hours per second of audio, but sounds really good. Polly can create audio faster than you can listen to it.
Surprised that Icelandic is being offered. As a native Icelandic speaker; to me it sounds about as good as can be expected with such a service. Isochrony (had to search the dictionary for that one) is a bit off but expected based on the context of phrases / words used to create the samples.
Just me, or do the foreign voices sound much more realistic than the English speaking ones do? It could be the fact that I am not a native speaker of Icelandic, French etc. so perhaps to a native speaker, it may still seem robotic and sterile, but to me, the inflections and cadence sounds much more natural in the non English synthetic speech.
The French sounds a bit flat to me but I could easily mistake it for a somewhat bored human reading in mostly a monotone. But I certainly don't know French well enough to be sensitive to unnatural pauses or odd inflections.
It really is quite good--even if I really wouldn't want to read an entire book read this way or would mistake it for a human. It definitely gets me thinking about ways to use this service.
This sounds better than OSX text-to-speech for audiobook purposes, but the 1500 character limit per API call is annoying. Instead of sending the ebook text in full, I have to split by paragraph and (occasionally) sentence and then stitch everything back together with manually inserted pauses, making the audio a bit uneven.
56 comments
[ 0.27 ms ] story [ 107 ms ] threadhttp://g-ecx.images-amazon.com/images/G/01/ivona/static//med...
I hoped there would be new & better german voices
English male sounds surprisingly bad. English female is better.
It'd be tolerable to hear a voice-interface with this, but it'd be maddening to try to listen to a book this way.
> F. Voices. Subject to the terms and conditions of this License, you may use the system voices included in the Apple Software (“System Voices”) (i) while running the Apple Software and (ii) to create your own original content and projects for your personal, non-commercial use. No other use of the System Voices is permitted by this License, including but not limited to the use, reproduction, display, performance, recording, publishing or redistribution of any of the System Voices in a profit, non-profit, public sharing or commercial context.
Which truly sucks, because they don't even give you the option to pay for such use.
https://d0.awsstatic.com/product-marketing/Polly/pt_br_vitor...
Unfortunately, I agree with Mizza regarding the quality.
edit: ah it was just the german version that is not available any more, english one seems to be still the store: https://play.google.com/store/apps/details?id=com.ivona.tts
You can read this as: We're pleased to announce our really cool TTS feature that took a lot of engineering know-how and effort ... but you'll have to click through, because we can't seem to get around the limitations of our CMS to embed audio content in a blog post.
EDIT: oh, I had no idea they have used Ivona
https://cloud.google.com/speech/
Computer generated voices feel most robotic when their intonation of a word is abnormal or their pauses between words make the sentence feel choppy. The intonation and natural pauses between words is very good for all of the main voices.
The Japanese voice Mizuki was the most comical addition, since I can't think of a real situation where she would ever actually be used. Mizuki speaks Engarish (the Japanese version of English) beautifully, but any Japanese person who can understand Engarish will also understand English. Also, Mizuki doesn't add the correct vowel ending to all words, e.g. she correctly says "cheezu" for "cheese", but says "steku" instead of "steki" for "steak".
My impression is that it's actually quite common for Japanese people to understand Japanese accented English more easily than a native English speaker.
>she correctly says "cheezu" for "cheese", but says "steku" instead of "steki" for "steak".
"Suteeku" sounds like what a Japanese person who knows English well but has a strong accent might say. "Suteeki" is more of a corrupted loanword.
This is actually very true, made me think back to many situations in Japan where adding a strong Japanese accent to my English words made it comprehensible to the listener (just as adding a Japanese accent to the generated speech makes it nearly incomprehensible to a non-Japanese speaker).
My biggest problem with CereVoice, though, has been its terrible web API. It doesn't support streaming output, so it renders the audio to an Amazon S3 bucket and then returns a URL, which is pretty inconvenient (and slow). You have to do the same for transcripts, too. So, if you want everything, you have to make 3 separate HTTP requests and parse 2 XML documents for one round of synthesis.
IBM Watson's TTS API gets it right, imo. Its streaming mode returns audio frames and transcripts over a WebSocket connection.
I realized this tool might not be cheap, since it may take the voice actor/actress 2 hours per day to produce my content (2-hour driving commuting per day for me). To get familiar local accent, it costs ~$36 in Australia, and maybe slightly cheaper for US accent. The value it brings me can hardly justify the cost.
Now, with Polly, things changed - it produces reasonable voice, and 2-hour content would only cost ~$0.3. I decided to launch my service as soon as Instapaper approves my API request.
At the same time, put your email here: http://readlater.launchrock.co/
EDIT: It is robot-like and nowhere near the quality of the samples that Amazon provided, but it gets the job done.
https://deepmind.com/blog/wavenet-generative-model-raw-audio...
This could get a bit confusing for some folks.
It really is quite good--even if I really wouldn't want to read an entire book read this way or would mistake it for a human. It definitely gets me thinking about ways to use this service.