Program that Turns Audio Clips into Pimsleur Clips?

[Xenops]

I thought I remember seeing a thread about a program where you can open some audio, and it automatically breaks up sentences like you have in a Pimsleur lesson . Does anyone remember what this program was?

Thank you.

[tangleweeds]

Perhaps you're thinking of the app that user ryanheise was making?

[iguanamon]

You're thinking of Gradint.

Gradint is a program that can be used to make your own self-study audio tapes for learning foreign-language vocabulary. You can use it to help with a course, to prepare for speaking assignments, or just to keep track of the vocabulary you come across.

The method: Gradint uses a variant of the “graduated-interval recall” method published by Pimsleur in 1967. It’s like audio flashcards that appear in a special pattern designed to help you remember. The Pimsleur accelerated language courses use several techniques (they say some are patented), and Gradint does not imitate all that, but this particular 1967 idea is now in the public domain so Gradint can use it to help you learn your own choice of vocabulary.

The page includes setup for a windows machine... but... how are your computer skills?

[lingua]

Audacity will do this but the process is a bit tedious. I've only done it with mp3 files though so I don't know if it has other capabilities that you might need.

[Xenops]

Perhaps you're thinking of the app that user ryanheise was making?

Mmmm no, but thanks for looking.

You're thinking of Gradint.

Gradint is a program that can be used to make your own self-study audio tapes for learning foreign-language vocabulary. You can use it to help with a course, to prepare for speaking assignments, or just to keep track of the vocabulary you come across.
The page includes setup for a windows machine... but... how are your computer skills?

Thank you, this is what I was looking for. They do have a version for Mac, and it looks like something I could handle--it just looks more tedious than I expected.

Audacity will do this but the process is a bit tedious. I've only done it with mp3 files though so I don't know if it has other capabilities that you might need.

I've also thought of using Audacity--maybe I'll try both.

[tangleweeds]

Ah, this is the one where I can never figure out quite what it *does*! Now far less surprising that this confessed app-addict couldn't bring it to mind.

Could a user please help me understand it better?

[Xenops]

Ah, this is the one where I can never figure out quite what it *does*! Now far less surprising that this confessed app-addict couldn't bring it to mind.

Could a user please help me understand it better?

Looking at the opening page, it looks like for every word you want, you have to put in a separate audio file. What the program does is create one long audio file that shuffles the words in a SRS fashion. You also have to have short audio files for the English prompts (or you can just record yourself).

[rdearman]

Looking at the opening page, it looks like for every word you want, you have to put in a separate audio file.

When you say "it" I assume you are talking about the Gradint program? Gradinit is nothing like Audacity or WorkAudioBook, it actually generates Pimsleur like lessons into MP3 or WAV.

I have used this for making Pimsleur like Chinese audio. You can use words but you shouldn't. You should be using sentences like Pimsleur. Gradint uses the same algorithm as the original Pimsleur. I generated about 20 CD's in both Mandarin and Setswana. I put a link below to Lesson 20 of the Setswana that I generated. Basically, you record the prompts into an audio file (eg "Please say...") or use the text-to-speech version if you don't want to record the prompts. Then it will use either audio files you've provided for the target language, or text-to-speech.

If you listen to the one I did below for Setswana, I took the words and sentences from the Peace Corps Setswana recordings and made Pimsleur like prompts then the system generated over 700 lessons between 20-30 minutes long. It uses the settings you give it for length of lessons, number of repetitions, etc. It is an amazingly useful program, although it is very fiddly to set up. However, once I understood the various inputs it was a snap, and the author (a Cambridge University Professor) is very helpful with supporting the tool.

It works out how often to introduce new words and phrases, and it uses the algorithm to repeat them to you periodically. If you wanted to, and you have some films, you could use EMK's substudy program (https://lib.rs/crates/substudy) to rip all the dialogue out of a film using the timing from the subtitles, then using the Gradint program to generate a Pimsleur course based on the film.

https://drive.google.com/file/d/1Vg-MjT ... sp=sharing

For my Chinese lessons I used the audio from SpoonFedChinese shared ANKI deck and generated something like 1000 lessons. (There is a lot more audio than for the Setswana Peace Corps lessons).

EDIT: I will not upload a link to those because there is some controversy surrounding the copyright of some SpoonFedChinese audio files. But one other thing. Once you've configured everything you can batch create the lessons. So I generated 1000 lessons in less than 10 minutes and I could listen to them in my car 30 minutes per lesson.

[Xenops]

Looking at the opening page, it looks like for every word you want, you have to put in a separate audio file.

When you say "it" I assume you are talking about the Gradint program? Gradinit is nothing like Audacity or WorkAudioBook, it actually generates Pimsleur like lessons into MP3 or WAV.

I have used this for making Pimsleur like Chinese audio. You can use words but you shouldn't. You should be using sentences like Pimsleur. Gradint uses the same algorithm as the original Pimsleur. I generated about 20 CD's in both Mandarin and Setswana. I put a link below to Lesson 20 of the Setswana that I generated. Basically, you record the prompts into an audio file (eg "Please say...") or use the text-to-speech version if you don't want to record the prompts. Then it will use either audio files you've provided for the target language, or text-to-speech.

If you listen to the one I did below for Setswana, I took the words and sentences from the Peace Corps Setswana recordings and made Pimsleur like prompts then the system generated over 700 lessons between 20-30 minutes long. It uses the settings you give it for length of lessons, number of repetitions, etc. It is an amazingly useful program, although it is very fiddly to set up. However, once I understood the various inputs it was a snap, and the author (a Cambridge University Professor) is very helpful with supporting the tool.

It works out how often to introduce new words and phrases, and it uses the algorithm to repeat them to you periodically. If you wanted to, and you have some films, you could use EMK's substudy program (https://lib.rs/crates/substudy) to rip all the dialogue out of a film using the timing from the subtitles, then using the Gradint program to generate a Pimsleur course based on the film.

https://drive.google.com/file/d/1Vg-MjT ... sp=sharing

For my Chinese lessons I used the audio from SpoonFedChinese shared ANKI deck and generated something like 1000 lessons. (There is a lot more audio than for the Setswana Peace Corps lessons).

Thank you for adding more detail, I appreciate it. My question: how did you configure it differently? I've played with the code on the cards fields for Anki, but I'm not seeing where I change anything for Gradint. Also, do I put an entire sentence in the word01, and other for word02, etc?



This was the only clear video I could find, but my French in beginner at best. I'm not clear on how Gradint can match an audio segment with a text word.

Thank you for your help.

[rdearman]

I'm a UNIX guy, so I rarely use a GUI if I can avoid it!

I just ran the script. But basically you need to configure everything first. So for example in the prompts directory there will be files like "repeatAfterMe.txt" and inside the file will be one line "Repeat after me." so Gradint will use this text as a prompt and use the sentence inside the file with text to speech to generate a prompt when required and add this to the MP3 it is generating. If you have a file called "repeatAfterMe.mp3" as well as the text, it will use your MP3 instead of the text-to-speech engine.

I created another subdirectory in the same place as the prompts (in my case in a directory called samples) which had the audio files I wanted to use, one in English, the other in Setswana. Example file names below.

985_you_have_a_book_en.mp3
985_you_have_a_book_tn.mp3

Gradint then uses these files to create the English prompt and the Setswana answer. Then I just ran a for loop:

Code: Select all

for N in 1 2 3 4 5 6 7 8 9; do python gradint.py "outputFile='lesson$N.mp3'"; done

This loop would generate 10 lessons, although I typically did "for N in {1..500}" to get 500 lessons. I don't think you can generate more than one lesson at a time with the GUI.

I am a little confused about why you are mentioning ANKI, since this program only generates a sound file where it has concatenated all the prompts and target language audio into a single lesson X minutes long. As far as I'm aware it doesn't do anything related to ANKI. The only reason I mentioned ANKI in the previous post was just because that is were I extracted my Chinese audio files from, because they were stored in the ANKI data directory.