Why Your Transcription Tool's Minutes Keep Shrinking: AI Credits vs Flat Pricing (2026)
If you opened your transcription tool one morning in the last year and found you suddenly had far fewer minutes than you remembered paying for — you're not imagining it. Across 2025 and into 2026, several of the biggest names in transcription quietly rewired how they charge. Generous minute allowances shrank. "Hours" became "media minutes." Flat features turned into metered AI credits that drain a little every time you touch a button.
This isn't a conspiracy — it's the economics of running large speech and language models catching up with the pricing pages. But it does mean the way you evaluated a transcription tool two years ago is out of date. This guide explains exactly what changed, how to decode the new pricing language, and how to pick a plan that won't strand you mid-project.
The Quiet Repricing of 2025-2026
Two changes tell the whole story.
Otter.ai cut its Pro plan from 6,000 minutes to 1,200 minutes a month — without dropping the price. That's a roughly 80% reduction in allowance. As of 2026, Otter Pro gives you 1,200 monthly transcription minutes (about 20 hours), caps individual conversations at 90 minutes, and — the part that catches people out — limits you to 10 file imports per month. If you upload recordings rather than transcribe live meetings, ten files disappear fast. Podcasters, researchers, and course creators routinely blow through that in a week and get funneled toward the pricier Business tier.
Descript replaced "transcription hours" with "media minutes" and introduced metered AI credits. After its September 2025 overhaul, Descript counts any audio or video you bring into the editor against your media-minute allowance — not just what you transcribe. On top of that, AI features (its Underlord assistant, Overdub, Studio Sound, and more) now consume a separate pool of AI credits with hard ceilings per tier. Credits don't roll over. A single Studio Sound pass can burn around 10 credits, so a plan with a few hundred credits is easier to exhaust than it looks.
The pattern is the same even where the details differ: the headline number on the pricing page no longer maps cleanly to "how much can I actually transcribe."
Three Pricing Models You'll See in 2026
Before you compare tools, learn to recognize which model you're looking at. There are three, and they behave very differently once you're at scale.
1. Metered minutes (the classic)
You get X minutes per month. Simple in theory, but watch for three traps:
- No rollover. Unused minutes vanish at the reset date. If you have a bursty workload — nothing for three weeks, then a conference dump — a monthly cap punishes you.
- Per-conversation limits. A 1,200-minute plan that caps single files at 90 minutes can't transcribe your two-hour webinar in one go.
- Import caps. The sneakiest one. You may have minutes left but be blocked because you hit a monthly file limit.
2. Metered AI credits (the new default)
Instead of (or alongside) minutes, you get a pool of credits that every AI action nibbles at — transcription, summaries, cleanup, voice tools. This makes cost genuinely hard to predict, because your bill now depends on which features you use, not just how much audio you have. If you can't answer "how many videos can I do this month?" from the pricing page, you're looking at a credit model.
3. Flat / per-item pricing
You pay a fixed amount per video (or an unlimited monthly rate) and every standard step — transcription, captions, exports — is included. Nothing meters down as you work. This is the model that got less common in the enterprise tools and more valuable for everyone who just wants predictable output.
There's a fuller breakdown of the raw dollar math in our 2026 guide to what video transcription actually costs and the transcription ROI and cost-savings analysis if you want to model your own volume.
How to Read the Fine Print (a 5-Minute Audit)
Before you commit to any transcription tool this year, run this checklist against its pricing page:
- Is the unit minutes, credits, or items? If it's credits, find the conversion — "1 credit = ? minutes of transcription" — and whether other features share the same pool.
- Do unused units roll over? Almost always no. Assume you lose the remainder every month.
- Is there a per-file or per-import cap? Look past the big number for a smaller one buried in the feature table.
- What's the longest single file you can process? Critical for webinars, sermons, depositions, long podcasts.
- Are exports (SRT, VTT, TXT) free, or do they cost credits/upgrades? Export gating is a common upsell trigger. Our SRT vs VTT format guide covers which format you'll actually need.
- What triggers a forced upgrade? Minutes, imports, seats, or feature access? Know which wall you'll hit first.
If a tool can't give you a clear answer to #1 and #2, that's not an accident — opacity is the point. Move on.
Who Gets Burned by Credit Models — and Who Doesn't
Metered credits aren't evil; they're just a bad fit for certain workloads. You'll feel the pinch most if you're:
- A podcaster or interviewer uploading long files. Per-conversation limits and import caps hit you first. (If that's you, our best podcast transcription software comparison benchmarks the honest limits.)
- A course or webinar creator with occasional but large recordings — bursty volume plus long files is the worst case for monthly caps.
- An agency or team running unpredictable client volume, where one busy month blows the budget.
- A researcher transcribing batches of interviews on deadline.
You'll barely notice the credit model if you transcribe a handful of short clips a month and never touch the advanced AI features. For everyone else, the safest bet in 2026 is a plan where the cost per video is fixed and visible before you upload.
The Flat-Pricing Alternative: What Tapescribe Does Differently
We built Tapescribe around the opposite philosophy: you should know exactly what a video costs before you upload it, and nothing should quietly meter down while you work.
- $1 per video, pay-as-you-go. No minutes to budget, no credits draining in the background. One video, one dollar, transcript and captions included.
- First 3 videos free, so you can check the accuracy on your own footage — accents, jargon, background noise and all — before paying anything.
- Plans when you scale: Starter at $29/mo (50 videos), Pro at $79/mo (200 videos), and Business at $199/mo with unlimited videos. Note the difference from a credit pool: a Tapescribe "credit" is one whole video, not a slice of shared AI budget that every feature eats into.
- Exports included. SRT, VTT, and plain text come standard — no export paywall, no format tax.
The point isn't that Tapescribe is always the cheapest line item (for a light, occasional user, a free tier somewhere may cover you). The point is predictability: when your plan is priced per whole video, you can't get stranded 25 days into the month because an invisible credit meter ran dry.
If you're specifically escaping one of the tools that repriced, we keep honest, feature-level breakdowns in our Otter.ai alternative guide and Descript alternative guide.
Frequently Asked Questions
Why did Otter reduce its minutes from 6,000 to 1,200? Otter cut the Pro allowance to 1,200 minutes/month in its 2025-2026 repricing while keeping the price the same, which effectively raised the cost per minute and pushes heavier users toward the Business tier. The 10-file-per-month import cap compounds this for anyone who uploads recordings rather than transcribing live.
What are AI credits in transcription tools? AI credits are a metered currency that certain AI actions consume — transcription, summaries, audio cleanup, voice generation. Because multiple features draw from the same pool and credits usually don't roll over, your real monthly capacity depends on which features you use, making costs hard to predict.
Is unlimited transcription actually possible? Yes, but usually as a flat monthly rate rather than a truly infinite metered plan. Tapescribe's Business tier ($199/mo) offers unlimited videos, which suits high-volume teams who'd otherwise spend their time managing credit balances.
How do I avoid surprise transcription caps? Run the five-minute audit above before subscribing: confirm the unit (minutes vs credits vs items), check for rollover, find the per-file and per-import limits, and verify exports are included. If the pricing page can't answer "how many videos can I do this month," pick a flatter model.
The Bottom Line
The 2025-2026 repricing wave didn't make transcription worse — the models are better than ever. It made pricing murkier. Minutes shrank, hours became media minutes, and flat features became metered credits that drain as you work. The defense is simple: know which of the three pricing models you're buying, read the caps that sit below the headline number, and favor predictability when your volume is bursty or growing.
If you're tired of doing minute-and-credit math, try Tapescribe free — your first 3 videos are on us, and after that it's a flat $1 per video with transcript, captions, and exports included. No shrinking minutes. No invisible meter.