Survey - what's your current go-to for meeting transcriptions?
Attempt number two to post this... Survey - what's your current go-to for ai meeting transcription? Given the proliferation meeting transcription apps, I'm wondering what people here are actually using and liking? I've tried several over the last year: Hyprnote (changed its name twice now to "Anarlog"?), Speakr, Transcribe X, Caption Snap and have some other tools that support transcription but I haven't tried because I can't really figure them out in general... (sorry, Alter) My biggest issue is that I meet with a lot of different people in a week and in many meetings it's not uncommon for there to be more than a dozen participants. I have yet to discover an app that nails speaker identification. For the ones that claim to support it, you only see them demo a 60-second meeting with 2 people. They don't showcase examples of a meeting with 10 people talking over the course of an hour, because that would show users how ill-suited their solution is for the kinds of meetings that we actually have and need to track. The closest I've seen to flawless speaker identification is Caption Snap, because speaker attribution is handled by the meeting platform itself as opposed to the transcription app; however this approach has its own set of issues and the output is currently not reliable enough to be trusted. So, for those of you that meet with more than one or two people at a time, and have nonstop meetings all day every day, what are you using to capture transcripts with reliable speaker identification? I genuinely don’t care about the money if the solution is rock solid.
已收集讨论
Macwhisper using Claude for summary works well with our morning meets with 8 people.
How are you managing disclosure? Do you say at the start of every meeting that you’re recording the meeting?
four out of five of my meetings are recorded by the organizer anyway, so disclosure is already there. The main drawback is often an organizer does not share access to the transcription or recording and even when they do, it’s a lot of manual steps to transpose the temporary data out of Teams into my own permanent note database. for the one out of five meetings that aren’t recorded by the organizer, I don’t mind telling others that I’m “taking AI notes” which are really the only artifact I bother to keep after the transcription has been processed.
Not really meeting transcription, but Granola is phenomenal.
Granola. Love the MCP as it allows me to bring my meetings into Claude, and then keep my knowledge base and PKM updated at all times
Hello Alter developer team here. What made you not try us? (Are we so complicated?) We do offer speaker identification both locally via Pyannote or on the cloud.
Hey, XSpeak dev here. While I can't say that speaker diarization is perfect in XSpeak, I improve it constantly and would be happy to know how it works in your scenarios and improve further.
I had the same problem. I was feeding a bunch of meeting notes into my second-brain setup, but the speaker names were wrong just often enough that it made the whole thing feel untrustworthy. Claude would mix people up, assign follow-ups to the wrong person, or treat a bad speaker label like it was fact etc. etc. That’s why I built Transcripted. I wanted a no-bot, 100% local Mac app that records meetings, transcribes them, and saves clean Markdown I can use with Obsidian, Claude, and Codex. The part I’m most focused on is speaker memory. Transcripted records meetings then captures speaker fingerprints, so after you identify someone after a meeting, it can get better at recognizing them in future meetings. Not magic, but the idea is that the more you meet with the same people, the less “Speaker 1 / Speaker 2” cleanup you have to do. It’s at transcripted.app. Free and open source. I’d genuinely love feedback if you try it.
I use Alter, but I do not have a lot of meetings currently, so it's limited more to calls. It's really not that complicated out of the box. But speaker identification is not flawless, and I've yet to try a solution that actually is. That being said, zoom has this built in, google meet has this built in, and microsoft teams does too. So I don't see why people are looking for third party tools for meetings through these platforms unless it's disabled for them.
Thanks. I did look at Alter's YouTube channel before posting to see what's new since I last checked it out and as of two days ago, they're promoting Speaker Suggestions. But the demo shared was very much "here's me and one other person chatting for a couple minutes. Watch me set Speaker 1 and Speaker 2" then doesn't go on to answer questions like "do I have to do this for every meeting or will Alter remember voices and automatically apply them in past or future meetings?" I can't bother with any app that makes me replace "Speaker 1" with a named person (even if the speakers are in an easy to select list of names) if I have to repeat that exercise for every meeting. Basically, if I say "this person talking right now is John Doe" then I need it from that point forward to recognize John's voice and apply it to all transcripts. Maybe Alter does that, but it wasn't clear in the demo. Also, seems like Alter is locked into Apple apps like Mail and Calendar? My work is exclusively in Outlook so I don't think I can leverage lots of Alter's features since they're tied to Mail. That said, I'm probably a stupid person because Alter has never clicked with me. Same for Fluent and lots of other "all-in-one" ai wrappers.
I will give Alter another try but good to know not to expect perfection. Agree that the built-in tools (we use Teams) are actually really good; but too often I'm not the meeting organizer and the organizer doesn't share the transcript afterwards or enable transcription to begin with (I respect wanting privacy but work meetings are not private IMO - I'm not looking to secretly record others, I really just want to have a list of bullet points from every conversation I have so I can add them to my notes and search/reference those notes in the future).
TwinMind works pretty well. Just not on VC I think.
vowen
My Mac is a bit older, so good local models are kinda slow. I started using Instanttranscriber, it's a web service and has good speaker annotation
Snaply is an awesome free option.
Circleback for me. MCP makes it worth it.
noota works real well for me
meetjamie.ai
Granola
I tried to use Claude but couldn’t figure out the connections for stuff cuz im a dummy. I use otter and the built-in zoom note taker. I tried using macwhisper but again - dummy over here. I need tools that just work… otter is very expensive but it really does just work. I haven’t found anything else that is as simple to set up and use right out of the proverbial box but I’d be very open to any suggestions that anyone might have to share!
Probably not a popular option but I use the meeting transcriber built in to Notion. I already pay for the AI plan so I get it included and having a database of meeting notes in the app I use for tracking notes and doing pretty much everything else is incredibly useful. Basically means I can link meetings (with notes, attendees etc) to projects and then use the AI to summarise it all. So I get it to look at previous meetings on the topic of an upcoming meeting to give me meeting prep notes, then record the meeting so that the next one includes the contents of the most recent meeting in the note. Then I've got an agent that runs every night that goes through all the meetings from that day and updates a "Latest Update" section in my project note (that's the page relating to the project in my projects database) so I can see at a glance the latest position of every project I'm working on at a glance. The down side, as you can probably tell, is that it doesn't do any of that out of the box. It needs a lot of setting up and configuration of instructions and agents and you need to be invested in using Notion for everything. But it's so good when it's all working as intended.
I’ve been using fathom but have been peaking at granola because of the simple notes feel.
same frustration here. speaker ID is the one thing nobody seems to actually nail. Circleback has been the most reliable for me with larger groups but it's still not perfect. the 10+ person meeting is basically unsolved territory
I built an app using whisper and a local model to give me exactly what I want from my meetings (prospecting, sales calls, stakeholder meeting etc) everything integrated to an obsidian vault for ai memory and follow up email prompts as well. Probably could’ve got an off the shelf one, but this was fun to build and now it’s free.
If you want speaker identification you really need a bot to join the call. For botless, I’ve been moderately impressed with Tana over my previous app I was paying for but I have yet to land on one I love. Tempted to give granola and shadow another go but I can’t recall the reason I wound up not choosing them