The one supported format where the size limit is a real obstacle — and the one where converting before upload is worth doing.
You can upload WAV files without converting them, but the arithmetic is worth knowing: WAV is uncompressed, so an hour of 44.1 kHz stereo is about 635 MB and just over 8 hours fit in the 5 GB limit. For anything longer, convert to 128 kbps MP3 — transcription accuracy will not change.
Drop a file here or choose one from your device. We will show you the beginning of the transcript first.
skryba.ai transcribes on Scribe v2, which gets 2.2% of words wrong in Artificial Analysis' independent index — the lowest of every model measured. Gemini 3.5 Transcribe gets 2.6% wrong, Whisper large-v3 4.1%. It is the same model that writes the transcript preview you see before paying.
| Model | WER |
|---|---|
| skryba.ai (Scribe v2)the model we transcribe on | 2.2% |
| MAI-Transcribe-1.5Microsoft Azure | 2.4% |
| Gemini 3.5 TranscribeGoogle | 2.6% |
| Universal-3 ProAssemblyAI | 3.1% |
| GPT TranscribeOpenAI | 3.3% |
| Whisper large-v3OpenAI | 4.1% |
| Nova-3Deepgram | 5.2% |
Source: Artificial Analysis, AA-WER v2 index, non-streaming mode — read 27 August 2026. Seven of the 40-plus models measured: the top of the table and the names you already know. The index is measured mostly on English audio; for other languages, Polish included, no comparable public ranking exists. That is why we show you the start of every transcript before we ask for anything.
WAV stores raw samples, so its size follows directly from the recording parameters and does not depend on what is on the tape. The formula is simple: sample rate × bit depth × channels.
| Parameters | Typical source | Per hour | Fits in 5 GB |
|---|---|---|---|
| 16 kHz, 16-bit, mono | dictaphone, call recording | 115 MB | ~46 h |
| 44.1 kHz, 16-bit, mono | single-microphone recorder | 318 MB | ~17 h |
| 44.1 kHz, 16-bit, stereo | CD standard, default export | 635 MB | ~8 h |
| 48 kHz, 24-bit, stereo | studio session, film equipment | 1.04 GB | ~5 h |
Scroll the table sideways to see the remaining columns.
There is one rule, and it follows from how speech recognition works: the model analyses the speech band, not sonic fidelity. A WAV's surplus quality does not buy a single additional correctly recognised word.
If WAV is an archival requirement on your side, keep the original and send an MP3 copy for transcription. The transcript will be identical and the upload many times shorter.
At 44.1 kHz 16-bit stereo, just over 8 hours, since an hour is around 635 MB. At 16 kHz mono, typical of dictaphones, around 46 hours. This is the only supported format where the limit is a real problem.
At typical compression settings, no. The model works on the speech band rather than sonic fidelity, so recording conditions matter incomparably more than the format. A difference only appears at very low bitrates.
Export it to 128 kbps MP3 in any audio editor — the file shrinks roughly elevenfold and the transcript comes out the same. Do not resample below 16 kHz.
Yes, upload it as it is, the same as a WAV. Losslessness does not make the transcript any better, so convert only when you want a faster upload: 128 kbps MP3 gives the same text from a file an order of magnitude smaller.
