Transcription tools

What every tool here has in common

Every page above runs the same pipeline, so choosing between them is about the guidance, not the capability. One engine handles many spoken languages and reads speech straight out of audio and video containers alike. Finished transcripts open in a synced editor where clicking any line replays that exact moment of the recording, which is what turns correcting a misheard name into a two-second job instead of a hunt through the audio. Exports cover TXT, DOCX and PDF for documents, plus SRT and VTT for subtitles.

The limits are the same everywhere too, because they belong to the plan rather than the tool. The free plan covers 60 minutes a month with files up to 30 minutes each, exports as TXT and asks for no card. Paid plans raise the monthly pool and the per-file ceiling, open up the document and subtitle exports, and add speaker labels. On every plan, uploaded audio and video delete themselves after 30 days, and nothing you upload is used to train AI models.

How to choose: by file format, or by the job

The format pages (MP3, MP4, WAV, M4A, FLAC, OGG, MPEG, AVI) exist to answer the question every uploader has: does my file work as-is? The answer is almost always yes: speech is read directly from audio and video containers alike, so converting a file before transcribing it is nearly always a wasted step. Each page covers its format's real quirks: where iPhone voice memos live, why a browser can't preview an AVI it can still transcribe, what WhatsApp voice notes actually are.

The task pages (podcast, interview, lecture, meeting, dictation, AI summarizer) start from the job instead: turning an episode into show notes, a lecture into study notes, a meeting into minutes, or a long recording into a summary you can verify line by line. The same engine sits underneath (multiple languages, speaker labels on paid plans, word-level timestamps) with different guidance on top.

If your file is unusual and your task is ordinary, start with the format page. If the file is ordinary and the task is specific, start with the task page. Both routes end at the same upload box, so the choice costs you nothing.

What happens after you upload

Processing runs faster than real time. A one-hour recording is usually ready in two to five minutes and short clips finish in seconds. Jobs run in the background with automatic retries, so you can queue several files and collect them from the dashboard together rather than waiting on each one.

The first thing worth doing with a finished transcript is a short pass over the names. Speech recognition is strongest on ordinary conversational words and weakest on proper nouns, so a surname, a product name or an unfamiliar place is where corrections cluster. Typing those terms into the vocabulary box before uploading, which is available on every plan, prevents most of them, and the synced editor catches the rest. Numbers deserve the same glance: spoken figures are converted to digits automatically, which is what you want in a transcript, but currency amounts and percentages are where a formatting slip is easiest to miss.