I decided this weekend to try an automatic transcription service to create private content out of some podcasts.
Does anybody have any experience with these? Which ones have you had success in? I’m using it for Greek which I understand is very phonetic so perhaps I have more options.
I’ve not used any personally. I believe Steve has mentioned Happy Scribe in his videos. Looks like there’s a couple of pricing options. Automatic or Professional.
Found this review. Sounds promising, plus there’s an editor that highlights areas where it’s not sure it got the transcription right. Again, I’ve never tried it myself.
Ah yes, thanks. The pains of studying a language with little intermediate content 
Anyway, it’s all part of the process.
Never done it myself, but I’ve heard from other users that you can just take the audio file, upload it to youtube, add auto-generated subtitles, publish (privately I guess) the video and then use the importer add-on to get it on LingQ.
I have no idea about the resulting quality of the transcription, probably not very good but you could give it a try.
If I get permission from the authors, I’ll try and share. I’m excited to transcribe: Γνωρίζοντας την ιστορία μας series by ΣΚΑΪ and the LIFO Podcasts. Not everything of course, that would be too expensive. Maybe I should pick a series and transcribe one a week to keep things fresh.
One thing I can’t stand about youtube autogenerated subtitles on other’s videos is there is never any punctuation. Yes, one can mostly figure it out as they read, but I find it so unappealing I don’t even bother to try and import those videos. I assume this would be the case with the method you’re describing, but I’ve not tried it to be sure.
No idea whether this Happy Scribe I linked to does a better job with that, even on their auto generated version.
If you try it, let us know how it works. I think there is a bit of “free” access. If I can get some free time I may give it a shot myself. Fortunately, I have intermediate content coming out of my ears right now so I’ve not so far had much need to try something like this, but it might be nice for those times where I find something really interesting on youtube but all I get are autogenerated subtitles…youtube’s without punctuation, bleh.
You are correct, there is no punctuation at all, and yes, it is a hassle to edit, sadly. I did that with a couple videos as an exercise some time ago, with another program that autogenerated subtitles. It took me around 2 hours for a 7 minutes video (maybe more). Plus I got some native to proofread it, which added to the work.
I also tried the free version of Happy Scribe. I managed to autogenerate the subtitles for one video (10mins) in Japanese, and then I had to pay (I just checked again, for less than 25 hours bought it’s 12euros/hour, so for the podcast that RJDavies wants, one random series I found is 5hours ± so 60 euros. I let you guys decide if it’s expensive or not). The quality was honestly meh, I couldn’t see the difference with the autogenerated subtitles that you can find in YouTube. I unfortunately deleted the transcript, otherwise I could have showed it to you.
Japanese I would imagine might be very difficult for it to auto generate. I wonder how a more “Western” language stacks up. If I get some free time I’ll see if I can try it out.
Yeah, maybe it was just an isolated bad experience. You can log in with your Google account, if I remember correctly it didn’t take much time. The video I chose was a random one in YouTube that had autogenerated subtitles so as to compare it afterwards, although I didn’t end up going into detail because the first impression was not that good honestly. But I’m curious about your results, please let us know.
I wish! Share if you find please and thank you.
Some more thoughts… I don’t know much about auto-transcription services so I may experiment with a monthly subscription with unlimited transcriptions. A fellow user ( ericb100 ) and I think Steve too have suggest/used Happyscribe which is a one off transcription service. Probably better if you only had a few pieces of audio content in mind. I think I’d be more interested in a subscription service that gave me unlimited transcriptions allowing me to create content out of entire podcast shows. Another thing to bear in mind is how well the service handles foreign language (i.e. not English) and whether the quality differs for certain languages. At my level I feel punctuation isn’t that important; I mainly want to use the transcripts as a quick reference to keep my listening comprehensible. … I’ll do some digging around and post again in a few days!
Update
Decided to try Trint, they have a monthly (unlimited) transcription service which seems ideal as I’ve found a lot of podcasts I’d like transcribed. I value quantity of transcription over quality. The upload/transcription took around 15 minutes and I have to say the quality isn’t too bad. I tested it with some audio with background music. The AI capitalises the wrong word occasionally and misses a full stop but being an intermediate learner, I don’t need that level of accuracy - I can fill in the gaps myself. If you’re learning a language with a particularly difficult script, I’d imagine you’d want more accuracy.
I don’t know the optimum way to upload my transcripts, I remember Steve using the import eBook feature. For now I’m contempt uploading the script and audio separately.
HappyScribe usage update.
Actually, I’m very impressed with just the auto transcript they have. I had it transcribe a Deutsche Welle news segment that only had auto captions on youtube. Of course youtube’s are all lowercase and no punctuation.
Happy Scribe had punctuation, including commas. Uppercase on nouns. I didn’t comb the entire transcript. I think there were a couple of errors here and there, however, what’s great is that they have an “Editor” area. When you open it, it presents the transcription and has words and phrases that it feels it has questions on its own transcription. It seemed based on what I saw that most of these were correct. Regardless, you can update the transcript if you feel it’s incorrect. To help, it has the video/audio that if you click on the word or section it pinpoints that in the audio and allows you to play that section, including rewind or fast forward 5 seconds. So you can go over that section as many times as you want to try and understand what may being said. As the audio/video is played it highlights where you are in the transcript which is really nice too.
It looks like it tries to break off certain sections of the transcript that it may feel are a new scene or different speakers and allows you to enter the speakers names or whatever you may want to call them. In the particular video I transcribed there was a section that was somewhat of an “add” that had many speakers one after the other. Unfortunately it didn’t really get this correct and jumbled them as one speaker section…Not a huge deal I don’t think. There may be a way to break a section or paragraph apart, but not sure.
The other cool thing in the editor is it saves the revisions you save off. So you could go back a few steps potentially if you feel like you did something wrong. Or all the way to the beginning.
Your first “auto-transcribe” appears to be free and then I think you have to pay after, so it’s worth someone giving it a try to see if it’s something they would find useful.
From that point it looks like you buy “hours”. At certain ranges of pre-bought hours it discounts. So if you buy between one hour and 25 hours it is $12/hr. If you buy 26 to 50 hours it is $11/hr. For 50 and above it is $9.6/hr.
There is also the “professional transcription” service where an actual person performs the transcription. I did not try that as it costs quite a bit more.
Anyway, I’m pretty impressed. I suspect it does pretty well with some of the popular western languages. rafarafa indicated he was not particularly impressed with the Japanese transcription it did. I suspect many of the Eastern languages it may have more difficulty with, but maybe it’s also improved since he last tried.
If it’s short enough you may just be able to create a lesson and cut and paste the transcript. Or if it’s youtube, go ahead and import and then copy and paste over whatever transcription came in. If it’s less than 2200 words or so it should fit. Otherwise you may need to do the import ebook option