Skip to content
Notifications
Clear all

I compared Sembly's transcript to Rev's. Sembly was 92% as accurate for 1/3 the cost.

18 Posts
18 Users
0 Reactions
4 Views
(@hellerj)
Estimable Member
Joined: 2 weeks ago
Posts: 91
 

Absolutely. The "two parallel pipelines" overhead is real, but it's more than just vendor management. You also create a training and expectation problem for your team. They now have to learn two different review workflows and output quirks, which slows adoption and introduces friction into the very process you're trying to streamline. That's often where the real cost hides.


Trust the trial period.


   
ReplyQuote
(@devops_not_grunt)
Reputable Member
Joined: 5 months ago
Posts: 188
 

You're right about the cascade, but you're assuming a broken data lineage is even detectable. If your dashboard query for "Kubernetes" returns zero results, you just assume it wasn't discussed. The failure is silent.

That's the real cost of the cheaper service - not the review labor, but the decisions made from incomplete data. You can't validate what you don't know is missing.



   
ReplyQuote
(@brookel)
Eminent Member
Joined: 2 weeks ago
Posts: 31
 

That's a really solid real-world test, thanks for sharing it. I'm in a similar spot looking at these services.

The 92% figure for 1/3rd the cost is exactly the kind of trade-off I'm willing to make for internal team meetings. Honestly, for my use case - just getting the gist and action items for folks who couldn't make it - that's probably good enough. The high-end accuracy feels like overkill unless it's for something official.

Did you notice if Sembly struggled with speaker diarization at all? Like, keeping track of who's talking when people jump in? That's my biggest worry with the lower-cost options, more than a few word errors.


Self-host or die trying.


   
ReplyQuote
Page 2 / 2