If you found my video helpful, maybe also check out this one by Chais Meyer. In his automation, he creates a mini blog post and a Linkedin post from his transcription: ua-cam.com/video/eRJSXiXv8oE/v-deo.html
Thanks for the video, Philipp 😊👍 Based on your more recent video about string splitting, if you use 0CodeKit, you could produce a transcription service more cost effectively by using asynchronous python coding. OpenAI developed Whisper as open source software, so it can be used from a python library within your own code & without charge. It is also possible to split audio files larger than 25MB in to smaller pieces & concatenate the individual transcriptions 👍 Viv
Thanks for this great comment, Viv! It's certainly more cost effective to run the Whisper transcription locally. And a python script to do this wouldn't be terribly complicated. With my channel, I am trying to show ways to do this with no code at all, to make it more accessible to more people. I'm constantly impressed what is possible with tools like Make for non-programmers.
Hey Philipp! Love your content. Can you make a similar flow using Whisper, taking the transcript info and create headings for the main points and create variables to be captured into a Google Sheet. That would be great!
Phillip, thank you for your tutorials. I find them extremely helpful, as I just getting started with Make and Bubble. Curious if there is a rapid API that could be inserted before the audio file hits whisper to cut it into smaller (25mb ) and then as it comes out of whisper to reassemble so one only has one transcription and one summary per audio file?
Thank you, very glad to hear you found this helpful! If the 25MB filesize limit is a problem, you could use the "Cut Audio File" action from 1001fx to create smaller audio files. It's built into Make. If your source audio file is a giant .wav format, you could also try to convert it to a smaller .mp3 file with a CloudConvert "Convert a File" action.
@@automate_all_the_things That's super helpful! I had the same questions. I was trying to transcribe a zoom recording and I kept getting the error of file size. Grateful for the content you created. Would you consider showing the process with a third party that will cut the file easily in a make automation? Many of the tutorials are showing transcription with short files.
very nice, but when running this again with a different file same as first sucessfull test, meaning same format on files it failes saying invalid format ? any suggestions
@@automate_all_the_things it`s smaller and same format, for some reason deleting the first test file and also delete google doc file it worked with a new file
Lieber Philipp - Vielen Dank für den Workflow. Im Video hast du selbst den Hinweis gegeben, dass die Begrenzung für transkribierbare Dateien bei 25 MB liegt. Häufig sind die Dateien jedoch größer. Zeig doch bitte einmal, wie die Einbindung weiterer Module in den Workflow aussieht, der auch größere Dateien zu einer brauchbaren Zusammenfassung bringt. Besten Dank im Voraus!
The Whisper API can only handle these file formats (mp3, mp4, mpeg, mpga, m4a, wav, and webm) so you'd have to convert a video file, or detach the audio. There is a service called Transloadit integrated with Make. They have a free tier and offer a "Converts a video to WebM" action. Maybe you could use that to convert your video to webm and then send that to Whisper.
Yes, the 25 MB file size limit of the Whisper API can become a problem. You could add a "Cut Audio" module from 1001fx into the scenario. Or convert a large audio into a smaller file format with a CloudConvert module. Is Zapier or some other option better at handling large audio files?
![3 Ways to Create PowerPoint Presentations with ChatGPT [for Teachers]](http://i.ytimg.com/vi/JmVWz2PFVA0/mqdefault.jpg)
If you found my video helpful, maybe also check out this one by Chais Meyer. In his automation, he creates a mini blog post and a Linkedin post from his transcription: ua-cam.com/video/eRJSXiXv8oE/v-deo.html
I was running against a wall because I didn't download the file in the 2nd step. Appreciate your content - works like a charm now.
Yes that took me some time to figure out as well 😃Glad you found this useful!
Thanks for the video, Philipp 😊👍
Based on your more recent video about string splitting, if you use 0CodeKit, you could produce a transcription service more cost effectively by using asynchronous python coding.
OpenAI developed Whisper as open source software, so it can be used from a python library within your own code & without charge.
It is also possible to split audio files larger than 25MB in to smaller pieces & concatenate the individual transcriptions 👍
Viv
Thanks for this great comment, Viv! It's certainly more cost effective to run the Whisper transcription locally. And a python script to do this wouldn't be terribly complicated. With my channel, I am trying to show ways to do this with no code at all, to make it more accessible to more people. I'm constantly impressed what is possible with tools like Make for non-programmers.
Great video!!! Thanks a lot!
Glad you liked it, thanks for your comment!
You are amazing. Great tut, perfectly explained!
Thanks! I'm glad you found this helpful!
Hey Philipp! Love your content. Can you make a similar flow using Whisper, taking the transcript info and create headings for the main points and create variables to be captured into a Google Sheet. That would be great!
hello could you help me understand what result I could get if I add the jungle scout api for Amazon sellers to my scenario
Phillip, thank you for your tutorials. I find them extremely helpful, as I just getting started with Make and Bubble. Curious if there is a rapid API that could be inserted before the audio file hits whisper to cut it into smaller (25mb ) and then as it comes out of whisper to reassemble so one only has one transcription and one summary per audio file?
Thank you, very glad to hear you found this helpful! If the 25MB filesize limit is a problem, you could use the "Cut Audio File" action from 1001fx to create smaller audio files. It's built into Make. If your source audio file is a giant .wav format, you could also try to convert it to a smaller .mp3 file with a CloudConvert "Convert a File" action.
@@automate_all_the_things That's super helpful! I had the same questions. I was trying to transcribe a zoom recording and I kept getting the error of file size. Grateful for the content you created. Would you consider showing the process with a third party that will cut the file easily in a make automation? Many of the tutorials are showing transcription with short files.
Also, will cutting the file effect the SRT file?
hello I had some ideas for automation can you tell me what you think of them
Sure, lets hear them :) You can also reach me via email: philipp@lazy-android.com
very nice, but when running this again with a different file same as first sucessfull test, meaning same format on files it failes saying invalid format ? any suggestions
Hmm, maybe the second file is too big? Whisper can only handle files up to 25 MB in size.
@@automate_all_the_things it`s smaller and same format, for some reason deleting the first test file and also delete google doc file it worked with a new file
Lieber Philipp - Vielen Dank für den Workflow. Im Video hast du selbst den Hinweis gegeben, dass die Begrenzung für transkribierbare Dateien bei 25 MB liegt. Häufig sind die Dateien jedoch größer. Zeig doch bitte einmal, wie die Einbindung weiterer Module in den Workflow aussieht, der auch größere Dateien zu einer brauchbaren Zusammenfassung bringt. Besten Dank im Voraus!
hey great video, would it work from a video added to the google drive, or would we need to have the audio extracted first? thanks
The Whisper API can only handle these file formats (mp3, mp4, mpeg, mpga, m4a, wav, and webm) so you'd have to convert a video file, or detach the audio. There is a service called Transloadit integrated with Make. They have a free tier and offer a "Converts a video to WebM" action. Maybe you could use that to convert your video to webm and then send that to Whisper.
Transcription does not have timestamp when using Whisper module, whats the workaround for this?
Make isn't good for large audio files > 25MB (Whisper limit) and long texts input. No chuncking option
Yes, the 25 MB file size limit of the Whisper API can become a problem. You could add a "Cut Audio" module from 1001fx into the scenario. Or convert a large audio into a smaller file format with a CloudConvert module. Is Zapier or some other option better at handling large audio files?
@@automate_all_the_things Large text is also an issue since the API can't process. Needs to chunk and combine.
@@automate_all_the_things Respell has a good template for splitting then recombining the audio files for that :)