Image Credit - John DaSilva @ Action Items Lab
We may earn a commission from links on this page.
If you are anything like us, you spend a large amount of time in meetings, these days largely online. And usually those meetings are back to back, with little time to recap meeting minutes and action items to ensure the alignment in your meetings translates back to actions after the meeting.
This is where the new breed of AI driven transcription software comes in.
But transcription isn’t just for online meetings, there are a variety of use cases that transcription software can be used for:
Face-to-face meetings with a conference microphone
Customer interviews
Recording your own speech practice sessions
John and I highly recommend OtterAI for most business and personal needs. It's feature rich, handles jargon well, has good integrations, a free plan, and reasonably monthly premium plans for business and power users. In this review I will go into depth on OtterAI, a powerful transcription and meeting management service to give you back time lost from manual summaries.
Table of Contents
Summary: Otter Transcription Review
Overall, John and I found Otter to be a quite powerful way of transcribing meeting notes, getting meeting summaries, and even post-meeting management to capture action items, enable meeting attendees to comment, even tag people that were not in the meeting.
In addition to online meetings, it can be effective at capturing face to face meeting conversations as well. In one instance we connected a conference call microphone in a meeting room to let the transcription software run, after which we could query the AI about conversation.
However, if you are expecting the platform to be a completely hands off post meeting for a perfect “send to the CEO” recap, this won’t entirely work. While we both found that it worked pretty well for capturing jargon and understanding technical conversations for marketing, legal and medical fields, there were still errors in some of the text and even who the text is attributed to, requiring some manual fixes.
Let’s start with some of the basic features so you can see if it will work for you.
Integration with Meeting Software
OtterAI integrates with a variety of different meeting platforms
Zoom
Sync cloud recordings
Show live transcripts in browser window
Google / Google Meets
Sync Calendar and Contacts
Microsoft / Microsoft Teams
Sync Calendar and Contacts
In addition, Otter.ai has apps which integrate with other platforms. At the moment, this includes:
Slack
Contacts, Conversations and Notifications
HubSpot
We're less clear about what this integration does as we were not able to test it - it states clearly says you can add your OtterAI data to the CRM
Dropbox
OtterAI Transcription Software Interface & Ease of Use
Overall, OtterAI has a pretty simple, user friendly interface to navigate
For the most part, there are only a couple of interfaces where you will spend your time.
There is an overview screen (below) where you can see your navigation.
Below you can see the list of meetings from our test account. To get into a conversation, just click it.
After you click on the conversation, you are brought to the main meeting management page:
The above screen is where the magic really happens - your transcription overview and meeting functionality.
Summary section
This section has a general call summary
But the most useful part is the “Summary Keywords” section. In this section you can click on the keywords to jump to the sections in the document
A playback section - by clicking on any text in the script, a playback of the recording will begin playing. You can easily pause or manage the playback with standard buttons
An Editing button (more on this below) to manually change the transcript
A slightly hidden hamburger section with the following functionality
Export: A transcription export with a bunch of options:
Format: TXT, PDF, DOCX, SRT. For those of you that don’t know, SRT is a subtitle file format usually used for video transcription
There are options to export the transcription with and without speaker names, timestamps and with or without takeaways
Rematch speakers: This is an automated function which will re-review the transcription to assist in correcting any misattribution of speakers to the audio. My experience with this was pretty bad. When I ran this feature, it actually attributed all text to the same speaker. John had a much better experience. So run a test on this yourself before you do it on an important meeting transcription
Sync text to audio: This is another automated function to re-check the transcription text to the audio
Transcription time management tools:
Change start time
Reset start time
Show relative time
The section in the top right which is one of the most interesting parts of OtterAI.
Summary: This is an AI generated summary, with an update of the conversation every 1 - 3 minutes. This summary can also be used to jump to the point in the conversation simply by clicking on the text.
Takeaways: Takeaways is essentially a summary of anything you have manually selected or commented on
Text Notes: You can add any text notes you want in this section
Highlights, Comments, Action Items: More on this is the “Meeting Transcription Collaborative Features” section below.
The other interesting part is the Chat functionality
In this part you can ask the AI to do anything related to your transcription. A bit more on this below.
OtterAI also has excellent apps for Iphone and Android devices, with the ability to synch your account across multiple devices.
Otter Transcription Accuracy
And on to the main functionality which you are interested in - how well does Otter actually transcribe your meetings?
Transcription Test Setup
What we did to test the transcription accuracy was to create 3 scripts which had heavy amounts of technical language in the following 3 professions:
Performance marketing
Legal
Medical
To do this we asked ChatGPT to come up with the scripts. And to make it as realistic as possible, we had it insert a lot of jargon, acronyms, technical language, action items, times for next meetings, etc., just like a real meeting would have. Granted, the scripts had an "AI feel" in terms of how it wrote them, and reading the scripts felt a little hokey, though it got the job done.
So How Well Did It Really Perform?
To evaluate the accuracy, we plugged the transcription text back into ChatGPT with the original text, and let AI do the evaluation.
Overall, while there were several superficial comments from ChatGPT (e.g. as we read the script, there would be natural stutters, “ah’s” & “um’s”, etc.), we found there was significant amounts of accuracy in the transcription.
The absolute biggest issue with the transcription was “misattribution” of the speakers. For example, if I was speaking it would attribute that to John. A bit more concerning, this often seemed to happen in the middle of sentences. Depending on the use case of your transcription, this could be a major issue or a non-issue.
In fact, for two of the tests it handled the technical language extremely well. However, for the medical transcription, it was less accurate. For example “myocardial infarction” was interpreted as “my their myocardial and fracked infarction”. The great thing about OtterAI is that you can go back and click on the text to hear the original recording, and there was a bit of stuttering at that time which led to the inaccuracy.
In all of the transcriptions, some specific things we noticed:
As alluded to above, one of the main things about transcription software is that it literally transcribes everything - ah’s um’s stutters, everything.
There were some misunderstandings, but in general the meaning of the conversation comes through quite clearly. Here is an example of a mistake:
Original: "I'll conduct a platform-wide performance analysis."
Transcription: "How conducted platform wise performance analysis."
Some acronyms were captured quite well, others not as well. It may be dependent on the first letter. For example, things like “CPC” and “ROI” were captured perfectly, but “AVP” was captured as “a VP”.
Soundalike words can be confusing for it. For example, it transcribed “Google meat” instead of “Google meet”
“And” vs “N” when in context of other letters. For example:
We were discussing the Windows snipping tool which requires you to type “Windows key SHIFT and S”, and it thought “and S” was “Windows key SHIFT N S”. But it actually understood the context, and capitalized the “SHIFT”
When saying “A and B”, this is captured as “ANB”.
FYI, it captures profanity very well - just something to keep in mind.
Transcription Punctuation
Overall, I would say that OtterAI did a mediocre job of handling punctuation. So if you need something precisely readable (either because you need customer language perfectly readable, or are trying to impress your boss), be prepared to do your own punctuation review.
Transcription Editing
OtterAI has editing tools within the transcription, enabling you to make changes or corrections. You can do any of the following:
Add any text you want
Change the speaker in any section
Combine paragraphs together
Meeting Transcription Collaborative Features
OtterAI Transcription Highlighting & Commenting
In addition to the editing of the transcription, you can use Otter as a full fledged meeting collaboration tool. Many of the features are similar to those in Google Docs, including commenting and tagging people in the notes.
It is quite easy to engage with these tools using the tools which come up when hovering over the text. Those tools include:
Like
Emoji reaction
Highlight
Comment
Add Action Item
Copy text
Link
OtterAI Transcription Commenting
You can tag anyone in your OtterAI organization, not only the people within the meeting itself, making it a great way to call attention to others about decisions or actions that may impact others.
In the below section, I highlighted a section, and am now in the process of tagging and commenting on that section. You can see that just by typing the @ symbol and the first letter of John’s name, I can tag him and then add the comment.
In addition, after this is done, these comments are then brought up in the “Takeaways” section in the top right.
OtterAI Transcription Action Items
Action Items can be conveniently captured with the popup when hovering over any text. You can then select the person to be assigned the action.
These action items then appear in the action item summary, when you can then add additional comments.
Otter Transcription Automation: AI Chat Function
Some of the best tools are in the deceptively simple “Chat” section in the top right side of the main interface page.
This is not just a chat with others section, here you can "chat" with the AI to query parts from the conversation. And wow - this is the most underrated part of OtterAI. The power to query any part of the conversation and ask for summaries is unlike anything you have done before.
In the chat box, you can add “@Otter.ai” to engage with the AI, and then ask your questions. One of the best parts is asking the AI to summarize specific sections and capture all of the Action Items in the conversation.
Overall Otter did a great job at detecting all of the action items from the meeting. The main problem was action item “attribution” - many of the action items were assigned to the wrong person. This would clearly be due to the same problem mentioned in the "Otter.ai Transcription Accuracy" section about the misattribution of the statements in the first place.
What this means to you is for your most critical action items you should capture these manually - don’t leave anything super important left up to the AI. But still, after a few tries (and probably a bit of prompt engineering to coax out more action items or content from the meeting), you should be able to rely on OtterAI to capture everything you need to move your projects forward.
Just make sure after the meeting the right actions are assigned to the right people.
But beyond capturing the action items for the meeting, this functionality has the power to change how you review all if your meetings to ensure tight collaboration and alignment.
Does Otter AI Transcribe in Other Languages / Does Otter AI Transcribe in Spanish
At this time, Otter AI doesn't offer direct transcription support for other languages, including Spanish. This means it cannot understand and convert spoken non-English languages, including Spanish, into non-English text directly.
How to use Otter AI in Spanish - Workaround
Otter AI can process Spanish audio transcription in two ways:
Transcription with Machine Translation: Otter AI can transcribe the audio, essentially capturing the sounds spoken. Then, it utilizes machine translation to convert that captured audio into text, but in English instead of Spanish. The accuracy of this translation may vary based on factors like audio quality, accents, and the complexity of the conversation.
Speaker Identification and Language Recognition: When encountering Spanish audio, Otter AI can still identify speakers and recognize that the language spoken is Spanish. However, the actual transcript text won't be in Spanish unless translated using the method mentioned above.
OtterAI Pricing
Otter has a range of pricing options.
Taking from the Otter site:
OtterAI Transcription Summary
Overall Otter is an incredibly powerful software for transcribing meeting notes and even managing overall meetings. We found the collaboration and AI tools to be far superior to the standard way of taking meeting notes, and a huge time saver.
Comentários