board_strategy_review.mp3
Sample
A real transcript, with nothing corrected by hand.
30 seconds of a recorded meeting, transcribed by the same pipeline your file would go through. The speaker labels and the timestamps were worked out from the audio; the words are exactly what came back. Press play and read along, or read it with the sound off — it is all here either way.
Done
Take it with you
These are the same exporters a paying customer gets, run against the same transcript. Nothing is watermarked, nothing is truncated, and you do not need an account to click one.
What this shows
Two people talking over each other’s sentences, in a meeting, with no script — which is the case that separates a transcriber that works from one that demos well. Nobody told it how many people were in the room, and nobody has touched the result since.
It also shows the arithmetic. A clip this short does not price by the hour: it lands on the per-file floor, so this one would have come to $2.00. The rate card is one number and one floor, and it fits in a sentence.
And one in a language nobody told it about
The upload form has a language box, and it starts on "Auto-detect (any language)". This recording went through on that default — nothing selected, nothing corrected afterwards. It is a public-domain Japanese reading, and the transcript under it is what came back.
Speaker A
これは私が小さい時に村のモヘというおじいさんから聞いたお話です。昔は私たちの村の近くの中山というところに小さなお城があって、中山様というお殿様がおられたそうです。その中山から少し離れた山の中にゴンギツネという狐がいました。ゴンは一人ぼっちの小狐で、下のいっぱい茂った森の中に穴を掘って住んでいました。そして夜でも昼でもあたりの村へ出てきていたずらばかりしました。畑へ入って芋を掘り散らしたり、菜種柄の干してあるのへ火をつけたり、百姓屋の裏手に吊るしてあるトンガラシをむしり取って行ったり。いろんなことをしました。ある秋のことでした。二三日雨が降り続いたその間、ゴンは外へも出られなくて穴の中にしゃがんでいました。
What it does not show
Being straight about the limits is worth more than another paragraph of praise. The meeting has 2 speakers and there is room for 5 colours, so it exercises fewer than half of them; the Japanese reading has one voice, so it shows the language and nothing about telling people apart. Neither clip is long, and neither is difficult audio.
A longer recording, a busier one, and a harder one would each show more. None of them may be staged to get there, so they arrive when there is a real recording to publish and not before. These two are what we have, and both are unedited.
Then try one of your own
A sample somebody else chose is still somebody else’s file. Your own recording, with your own accents and your own crosstalk, is the only test that settles it — and if the transcription fails, the authorization on your card is cancelled rather than captured, so it costs you nothing to find out.