Please improve the audio quality!
Let me start by saying that I really like the service and the idea behind duolingo. Many thanks for creating a great language learning resource! By far the worst part though, in my opinion, is the quality of the audio. It seems to be the output of some automated text-to-speech function, which is often incredibly crappy (I have confirmed this for multiple languages with native-speaking friends). For beginners especially, this is probably even detrimental to their ability to develop good pronunciation and understand native speakers. The slow-down button helps in some cases, but only very marginally. I realize text-to-speech is a cheap content alternative, but having real native-speaker audio would be the best way to improve the service and would make me much more likely to recommend the service to friends.
I'm glad it's not just me! It often results in me getting the answer wrong just because, even after slowing the audio down, I have no idea which of several possible words she's saying. Headphones help a little, but it's still not good!
+1 for this and I'm telling you: If you can't hire a real native speaker to read the lines, the Crowd Source audio that so many people dismissed in that discussion thread is the way to go. Incentivize native speakers and professional language teachers to contribute, and simply make them give information about what region of the country their accent is from, age, gender, etc. Let us vote the samples up or down. Then not only do we have a great selection of audio to choose from, but we would eventually be able to set audio preferences for things like "Female, Napoletano" to train our ears to hear better in the region we are planning to live/travel/converse.
hello my native language is Spanish and if we talk about the audio in Spanish, in my opinion is not bad and the contrary is very good but I can understand that learning a language can be very difficult, I say it because I try to learn English and sometimes not I can understand the pronunciation.
Just to be clear, the audio is not always totally unintelligible. The problem seems to be the way the individual words are spliced together. I just think it's bad for beginners to hear poorly generated/spliced audio.
Agreed. I think it would be very difficult to do this, sadly. It is awful to learn all the words the wrong way though. It really can mess you up for good, I think. I think that it's decent though, and if you watch shows and listen to music in the language, it will probably help.
I don't think it would be so hard to do it. I imagine that many of the users are working on beginner modules, which have limited vocabulary. At least to start, it would be easy to record native speakers on the beginner vocabulary where a lot of users are.
I've found that as I continue in more complicated levels, with more challenging sentences, it's harder to understand the audio, and when I write or translate what I think I'm hearing, I get it wrong. I hope having a native speaker narrate, it would be easier to distinguish contractions and pronouns spoken.
Yeah, I've had the same experience. The automated text to speech gets much worse for longer phrases.
It's unclear sometimes for me too, but usually I get it if I ask for a slow repeat. However! I would like to add that I'm on a medium to slow internet connection, and already the audio takes a long time to load at times. Higher quality will lead to larger file sizes...
My question is: do we want to start reducing availability to those on slow connections, typically sitting in less developed countries?
Yes, I agree. For French learners particularly, the slow-down button may help to understand single words but not to develop a good pronunciation. "Liaison" and "enchainement" get lost in that case.
The problem for me in French is there seems to be a lot of audio distortion which it makes it hard to distinguish sounds so for example "pomme" and "pain", which should be clearly different, can be indistinguishable. I've been watching a lot of French television recently and I find the words are much clearer there. In German I have no trouble because there is no distortion.
I would love better audio, and I'm sure there will be some native speakers willing to volunteer. Especially in German, where a lot of the pronunciation is a bit off.
For the most part, I hate how Er (He) sounds like Ihr (You) in the audio files. I can distinctly hear the difference when a native speaker speaks.
That bugs me too but the Duo German team seem to think it's fine. I keep complaining but the only way I can distinguish er and ihr is context. In real life they certainly sound different - I'm going on what I was taught previously and how my native German speaking brother-in-law pronounces them. "Wird" being pronounced as "wilt" is the other irritation that is only resolved by context.
I found this thread as a result of a problems in the quality of the Esperanto recordings. The Esperanto voice is a human. There's one or more of the following on at least half the recordings: feedback, acoustic echo, compression artifacts, voice activated recording, recorded over cellphones, microphone too close to mouth, no foam sibilance filter, other audio processing artifacts.
At some point the sound quality problem went away. Now it's back. I don't know why.
I agree. It's hard to understand what the speaker is saying, but, it would be difficult to change the audio quality. So far, I recommend clicking the sound button when the sentence is visible.
I generally put it through google translate and a (german for me) dictionary to see differences and get a better feel for what the correct pronunciation might be
got to agree with that even when slowed down some words sound a lot like others, then i try to work out the rest of the phrase I will plump for the one I think fits,sometimes I get it right others I don,t for example tu y su, duermes y viernes, ah well, we learn by our mistakes, having said that if you were a native speaker it might sound obvious what the word or words are? i don,t know but you do have a valid point.
Don't forget that when you'll be talking in real life, the audio will be even faster and you'll have less time to respond. The Spanish fembot is pretty fine, but it is clear it's Latin American Spanish.
I imagine the number of permutations of all the auto-generated grammatically correct sentences is ridiculously large. If you are serious about improving your listening comprehension you should practice with listening to LOTS of native media. You can watch spanish movies, listen to spanish podcasts, watch spanish youtube or use a service like LingQ.com if you want to pay a cheap price for something decent.
Your good suggestions are a must if a language learner has dreams of actually conversing with a native speaker. Lots of free podcasts are available on iTunes. Many have transcripts so even if you don't understand what is being said you can follow along and practice mimicking the pronunciation. Currently, DUO's audio is of such poor quality I don't rely on it to learn correct pronunciation. Even Google Translate does a marginally better job than DUO. Compared to English, Spanish is a breeze to pronounce. French, now that's a whole other story:) Perhaps a future enhancement?
I think the Spanish audio is ok but a lot harder to understand than if a native Spanish speaker was speaking clearly. I also use podcasts that have native Spanish speakers (from Spain and Latin Am.) and I can understand them better than this synthesised voice. Could it be that native speakers of Spanish who listen to the audio and think it is fine do so because they are much more familiar with the language and their brain sort of fills in for the missing quality? I do also find that once I actually know what exactly the sentence is then it actually sounds more correct than before I know, which to me seems to be my brain doing a similar thing.
I'm learning german, spanish, french and portuguese, and i would say that the german and spanish audio isn't that bad. it's actually good. but the french audio is really bad, sounds very automated, and the portuguese audio is just horrible. sometimes impossible to understand. it's just so bad. and i agree to the suggestion of making real native speaker audio. why not multiple accents audio from which we can choose our suitable one. but for now, Duolingo is still a great tool even with these audios.
I think that the audio is surprisingly good. Still, probably is an area Duolingo will improve over time.
The audio in Portuguese is particularly bad on this site. But it says Portuguese is in beta, maybe they'll fix it. I'm pretty sure that speakers of Portuguese do not sound like short-circuiting robots.
I have been learning French for awhile now (not online) and was attracted to this program as a way of practicing and improving. But at times the French accent (when you are meant to type something you've heard) is just ... weird. I am in the Uk and used to hearing French from France. Is the accent in the software Canadian? It would be great if, like in car navigation systems, if you could choose a voice/accent for the language you are learning.
French is fuzzy and distorted compared to Spanish. It takes away from the enjoyment.
I've been using the iPhone app for about a month and just got on the website today. My biggest issue is also the sound. Sometimes even if I use the slow down feature (okay I always use it because rapid fire Italian is just very overwhelming to me) I get the words wrong because I'm not hearing what is really being said. I've lost count of how many times I've confused 'r' with 'l' in exercises.
I keep getting questions wrong because I can't hear the difference between key words, like "je" and "tu"! I am not an auditory learner, but that makes it all the more necessary for me to have a good audio.
That's because, in French, many of the letters remain silent, particularly the ends of words get dropped. or you have several words forming a liaison (one unit of sound). In Spanish, each letter is pronounced (except h). so it's fairly easy to write out Spanish words even if you've never seen it before. That assumes that the learner knows the sound of each letter of the alaphabet.
It's hard to say, I'm using this site via Google tv, so my audio is thru the tv which sounds pretty clear, but I do have to put on best audio source, music, movie etc to get the clearest sound, but it helps when I use the soundbar also brings it clearer, but overall on right sound setting I have no problem. I'm learning German and just now had the word (Royal), well it is pronounced way differently than an english speaking person would pronounce it, so is it really the audio or the different language pronunciation? Again on my tv sounds OK to me tho.
I find the audio quality quite acceptable in German, though not quite up to native-speaker standards. I haven't noticed any misleading pronunciations, but occasionally there's an audio glitch which makes a word unclear. Italian is less good, but still comprehensible -- there's a bit of a "long-distance telephone line" feel to the quality, but I don't feel that this is a big problem for comprehension. I've noticed that the intonation can be off though, which is a problem in a language where intonation can be the only difference between a question and a statement! I'd be interested to hear from someone at Duolingo whether there are any improvements in the works.
I've just been doing a lesson (Italian) and although I'm using headphones, I had to skip a couple of questions because, even with the audio slowed down and repeated several times, I was totally unable to decipher the words!
The Italian audio is shocking. It sounds like she is brething in between each word which makes it incredibly difficult to answer. Every other langauge I have tried has been awesome and I love Duonlingo as a whole, it is just impossible to complete the Listening component of Italian with such poor quality in the audio.
The spanish audio is good for the most part, but it's been glitching for me recently. With some words, the robotic person won't pronounce all the letters like veía sounded like vea but when I refreshed the page, she pronounced it right. And sometimes they emphasize the wrong part of the word or don't emphasize any part like hacían instead of hacíííííían she said hacien no joke. I knew what she was saying, but she didn't say it right at all. And of course it's harder to understand the automated version because she slurs some words together and is kind of glitchy (at least for me). I even changed my language to English (because I like to practice Spanish from both sides) and the English audio is really bad too. I couldn't understand what they were saying and i got some parts wrong and it was my language! So I can understand how other people are having trouble understanding the english voice because that is some bad audio. It's better to listen to music and watch movies to help learn how the language is spoken. I sure hope no one relies on the robotic person to learn how to speak whatever language they are trying to learn.
I agree, also I found out that TTS doesn't imitate people speaking in real life. I have a friend from Uruguay and whenever he talks to me in spanish I can't understand a word because I'm used to TTS.
We should propose people to submit their pronunciations as well
I am a little hard of hearing and I find the slow button a great help. I have been told by a friend whose first language is Spanish that the speech is very good. I thank you for this fantastic opportunity to learn another language , free, with such ease.
My boyfriend started the English track (he's a native Spanish/Japanese speaker) and...my god, the quality of the English voice is just terrible. I'm a native English speaker and I just had NO idea what she was saying!!
Still losing hearts, because even after listening several times, through headphones, some words are a complete mystery. It's SO annoying! Doing Italian, and can often not distinguish between "i" and "il" or "un" (and it's so distorted)
I've found that the Spanish is okay and French tends to be muddled (for the languages I spoke already pre-Duolingo and was thus used to how they sound in real life), but the Portuguese voice is leading me to some definite problems. The speech synthesizer is obviously very good, but the splicing is not great and the audio quality is not fantastic—and for learning language, it seems like real people speaking is the way to go with audio, no matter how good the synthesizer is.
I don't know what language you're learning, but I notice I start understanding the Spanish computer-lady better and better every time. If you visit a Spanish language country you'll notice their speech is faster than the fast lady version, and that's AFTER you ask them to slow down!
I just joined Duolingo today and I must say the text-to-speech is a huge weakness for Duolingo. I missed more than a few quiz answers that I would have gotten had the enunciation been clear. Luckily I'm not relying on Duolingo to learn pronunciation but I'll still need to deal with the text to speech in order to pass the quizzes. Native, actual human voices would be a huge plus! I know this service is free, but it doesn't hurt to ask, right? :)
The sound quality in the Spanish proram is "fuzzy" and, therefore I frequently have to use the slow version to make out what she's saying. This subject has come up hundreds of times in discussion over an extended period of time. I think it's important. I think also that it would be helpful to have both male and female speakers and some from latin america as well. If the sound was clear and the speakers were varied it would add a great deal to the learning experience.
The female Spanish female voice often puts the accent on the wrong part of the word. This can be confusing for someone who is learning.
The audio for the Russian lessons is really bad too. Glitchy and often cuts out. Many times even the slow speed has words that are unintelligible. I report the bad ones but I don't get the impression from the forums anything is getting corrected. I think they're just very low quality recordings. I KNOW it's free but I would gladly pay a subscription for better audio. By the way, I experience the same audio issues on the android phone app so it's not a problem specific to my PC.
I notice this post is 5 years old now. I am a relative newcomer to duolingo (French). In new tree parlance, I am approx 60% progressed to Level 3. Unfortunately I am finding the pronunciation, in particular the female, and most frequently used, voice, is becoming a limiting factor to the benefit of duolingo. As I now live in Quebec, Canada I listen to professional announcers on CBC Premiere radio and their pronunciation is far superior to the duolingo "lady". Unquestionably, duolingo would be much better if they were to emulate professional announcers vs the muffled/muddled voice in, evidently now much prolonged, use. [Of course the native street level Quebecois hasty pronunciation is an entirely other matter; but not something duolingo can resolve]