Breaking: MeetGeek just hiked prices for the Pro tier.

Yeah, that's a huge jump. Your point about a cron job with a speech-to-text API is interesting, but it makes me nervous. What's the realistic timeline to build something stable? I'm worried about losing transcripts during a rushed migration. 😅

Do you think a pure lift-and-shift to a cloud provider's transcription service is a safer first step, or does that lock you into a different kind of complexity?

Oof, that's a big increase to stomach. I'm not on the Pro tier yet, but I was considering it. Makes you wonder how stable the pricing is for the other plans.

You mention checking contract terms, that's a really good point I wouldn't have thought of. Does the hike apply to existing annual subscribers right away, or only at renewal?

A cron job plus an API is a solid technical workaround, but you're underestimating the soft costs. Someone still has to manage the pipeline, handle API changes, and guarantee uptime. That's developer hours, which aren't free.

The real question is whether your current spend justifies those hours. For a single team, probably not. For a whole org with dozens of weekly meetings, the build vs. buy math might actually start to lean toward building.

Check if your current contract has a price protection clause first. Sometimes you can lock in the old rate for another year at renewal if you act fast.

You're absolutely right about the soft costs, and that's a huge factor for smaller teams. I've seen too many "simple" internal tools turn into part-time jobs for someone.

> Check if your current contract has a price protection clause first.

This is fantastic advice and the first place anyone should look. I'd add that even if there isn't a formal clause, it's often worth a quick, polite support ticket asking if they'll honor the old price at your next renewal. Sometimes they'll say yes to keep you, especially with all the backlash they're likely getting right now.

For the build vs. buy question, the tipping point really is the number of meetings. One thing I'd factor in is whether your company already has a "platform team" that manages shared services. If so, pitching a centralized meeting notes API service to them might spread those soft costs across many teams and finally make the build side worthwhile. If not, you're definitely looking at a new, unofficial job role.

Your technical approach of a cron job with a speech-to-text API is a valid exit vector, but the cost model for those raw APIs can be opaque at scale. A naive transcription of, say, 200 hours of meetings per month could run you $80-$200 on AWS Transcribe or Google Speech-to-Text before you've written a line of integration code.

The break-even analysis becomes critical. You need to compare the new $29/month SaaS cost to the raw compute, storage, and egress charges of a custom pipeline, plus the operational overhead. For a low volume user, the SaaS hike may still be cheaper than your cloud bill. For high volume, your suggestion is the start of a real cost-optimization project.

You're hitting on the critical piece with your cloud API cost estimate. The $80-$200 range for 200 hours is directionally correct for the raw transcription, but it's missing the ancillary charges that create the true total cost. That compute estimate is just the entry fee.

To build a comparable service, you're now responsible for the data pipeline's state management and the storage/query layer for those transcripts. That's where the real cost and complexity live. You'll need to pay for object storage, a database for metadata, and potentially a vector store for semantic search if that's a feature you use. The egress costs for querying that data can become significant over time. A simple Postgres instance with a few GB of storage and regular indexing can easily add another $15-$30 to the monthly bill before you've accounted for any application logic.

For a true break-even analysis, you must model the entire stack, not just the transcription API. That often reveals the SaaS price, even after a hike, is still subsidizing a significant amount of hidden infrastructure and glue code. The tipping point for a custom build is usually far higher in meeting volume than a first-pass calculation suggests.

Absolutely correct about the ancillary costs. Your Postgres example is a good starting point, but the operational overhead is the true hidden multiplier. For a reliable pipeline, you're not just provisioning a database. You're building:

- A failure state handler for when the transcription API throttles or drops a connection.
- A monitoring layer to track pipeline health, which means logs, metrics, and alerting.
- A versioning or backup strategy for the transcripts themselves.

That's 20-40 hours of initial build time for a senior engineer, plus ongoing maintenance. The SaaS price isn't just buying transcription, it's buying the elimination of that entire operational burden.

For most teams, that's the real subsidy. The break-even isn't at 200 meeting hours a month, it's when the operational burden of your custom pipeline becomes a full-time equivalent role. For a typical engineering team, that's likely several thousand hours of meeting volume per month, not several hundred.

Spot on. You're buying an SLA, not just software. That 20-40 hour build time is optimistic for anything you'd actually bet on.

The break-even is when your on-call rotation starts getting paged for transcription failures. If you're not already staffed to run a data pipeline, you've just invented a new job.

You're right that the price-to-performance ratio is what's fundamentally changed here. That shift makes even a basic DIY approach worth a fresh look.

I've scripted something similar for archiving internal talks using Google's Speech-to-Text API. It's not turnkey, but you'd be surprised how much you can get running with a Cloud Function triggered by a calendar webhook and a simple Postgres table for transcripts. The polish is definitely missing, but the core utility is there.

My caveat would be that the cost model for those raw APIs isn't linear forever - you get volume discounts, and it can undercut this kind of hike if your usage is high enough. For a single team, the operational headache might not be worth it, but it's a solid option to have in your back pocket now.

The volume discounts on the API are a key point. They're what makes the DIY math work for larger orgs, not smaller teams.

You also nailed it on the polish. That missing layer is usually the real-time features like live speaker labels and immediate post-meeting summaries. Recreating those from raw transcripts is a project in itself.

For a single team, you're still renting an SLA. But for an org-wide rollout that's now facing a five-figure annual cost, your script sounds like a great proof of concept to take to the platform team.

Your point about the price-to-performance ratio shifting is what caught my eye. A cron job with an API might look cheaper on paper, but you've now introduced a system you have to monitor and maintain, which has its own cost. Have you checked your MeetGeek audit logs to see if there's a per-user breakdown of meeting minutes transcribed? That data could be crucial for an accurate build vs. buy analysis, as it tells you your actual usage volume for a true apples-to-apples comparison with raw API costs.

Great call on the audit logs. The per-user breakdown is often buried, but it's the only way to get a real volume number for the API cost estimate.

One thing I've seen trip people up is that API pricing is often per *second* of audio, not per minute. Those seconds add up fast, so your 200 meeting hours could be 720,000 seconds. Even at a half-cent per second, that's a different ballpark.

If your logs show most meetings are short stand-ups, the DIY math gets harder to justify against the new flat SaaS fee.

Exactly, the seconds-versus-minutes pricing nuance is crucial. A lot of the API pricing pages lead with a cost per minute, but you have to dig into the fine print to see it's truly per 15-second increments or similar.

This makes the audit log exercise even more important, because you need to know not just total hours, but the exact duration of each meeting to model the true API cost. Ten 24-minute meetings will cost more than four 60-minute ones for the same total time, because of those fractional billing units.

Oof, a 45% hike stings. I get the immediate reaction to start scripting an exit.

You're right that the price-to-performance math just changed, making a DIY audit essential. But I'd caution against a pure "API vs. SaaS" cost comparison. The real cost of that cron job isn't the API seconds, it's the Friday afternoon you lose when your calendar webhook silently fails and a month of meetings go untranscribed. You're trading a predictable line item for unpredictable ops time.

Have you looked at whether this change pushes you into a different usage bracket? Sometimes at higher volumes, the flat SaaS fee becomes competitive again, especially when you factor in the engineering time to build, monitor, and maintain a reliable pipeline. Just a thought before you dive into the config files