what temi got right
temi's pricing model — pay per minute, no subscription, no minimum — fits how most individuals buy transcription. you have a file on tuesday, you don't have one again until next month, and you don't want a recurring charge in between. most of the rest of the category drifted toward subscriptions for revenue-predictability reasons, and the buyer who used to land on temi increasingly has nowhere to go.
we keep that model. pay per file. no subscription. no minimum.
what stopped working
three things, in roughly the order they hurt:
- accuracy regressed. users with multi-year tenure on temi report transcript quality declining over the past two years. clean studio audio coming back as "a garbled mess of odd phrases." common technical vocabulary substituted for nonsense ("Bayesian" → "Beijing"). the cleanup tax went from minor to significant.
- the .docx is the product. temi delivers a Word document. fixing speaker labels means find-and-replace through paragraphs of text. verifying a quote means opening the audio in a separate program and scrubbing the timeline. for any transcript that needs to survive scrutiny — court, citation, publishing — that's the wrong substrate.
- support evaporated. user reports of multi-day silence on billing tickets. no phone number for billing issues. an organization clearly running on auto-pilot.
head-to-head
what shipping looks like at launch.
cleanup time
- temi
- 25–40% of audio
- ours
- near-zero
the editor
- temi
- .docx in word
- ours
- click word, hear audio. fix speakers in bulk.
subscription
- temi
- none
- ours
- none. pay per file.
audio leaves your laptop
- temi
- yes — uploaded
- ours
- no — runs in your browser
| temi | audiohighlight | |
|---|---|---|
| cleanup time | 25–40% of audio | near-zero |
| the editor | .docx in word | click word, hear audio. fix speakers in bulk. |
| subscription | none | none. pay per file. |
| audio leaves your laptop | yes — uploaded | no — runs in your browser |
where we're better
- cleanup time. our target is under 5% of audio length on the published benchmark corpus. temi's reported median is 25–40%. on a 30-minute interview that's the difference between a working hour and a working morning.
- browser workspace, not a Word document. click any word; that exact second of audio plays. flag a speaker label once; it propagates. highlight a quote; export it as a citation with a timestamp.
- custom vocabulary. per-account terms list. proper nouns and technical words stop being the recurring tax they are on temi.
- private mode. for therapy, legal, journalism, or any audio you don't want uploaded — see private transcription. temi has no equivalent.
where temi is still fine
- cleanly recorded single-speaker audio. a clean studio podcast monologue gets through temi's worst failure modes. the cleanup tax is real but bounded.
- for users who already have the workflow built around it. if you've spent two years training your workflow to expect a temi .docx, switching has a cost. if you're early enough that switching is cheap, it's the right time.
pricing comparison
temi: $0.25 per minute, no subscription. $7.50 for a 30-minute file.
audiohighlight: $0.25 per minute — pay per file, no subscription, no minimum, all features included. $7.50 for the same 30-minute file. private mode and cloud mode are the same price — no premium for on-device. 20% under temi at every file size.
switching
there's nothing to migrate. temi files are .docx; you keep them. transcription history isn't portable in any tool. the test case is to drop the same file into both and compare — which is what the benchmark is for.