Where your audio is processed
AssemblyAI through its EU endpoint (the audio is deleted from AssemblyAI as soon as the transcript is saved), Whisper local on Hetzner DE as fallback, and the audio in an R2 bucket with EU jurisdiction.
Hybrid transcription with AssemblyAI EU endpoint as primary and Whisper on our own server in Europe as fallback. 8 Spanish regional variants. ES↔EN code-switching.
AssemblyAI through its EU endpoint (the audio is deleted from AssemblyAI as soon as the transcript is saved), Whisper local on Hetzner DE as fallback, and the audio in an R2 bucket with EU jurisdiction.
Iberian, Mexican, Argentine, Colombian, Chilean, Venezuelan, Peruvian, Uruguayan: each fine-tuned, not lumped as "Spanish".
Per-segment language detection. If your meeting mixes Spanish and English, the transcript does too.
Who said what, with names you can edit. Bundled in every paid tier, not gated to Enterprise.
AssemblyAI primary + Whisper sidecar fallback. Outage of one doesn't stop transcription.
/benchmark publishes our WER per Spanish variant. Our model is not a black box.
Primary: AssemblyAI Universal-3.5 Pro via the EU endpoint (api.eu.assemblyai.com). Fallback: faster-whisper large-v3-turbo on our dedicated Hetzner DE server. Both run with multilingual support including ES↔EN code-switching.
7–9% on Iberian Spanish (vs ~14.7% on generic Whisper). Per-variant benchmarks are public at /benchmark with sample audios and methodology.
Optionally. You can keep audio in our EU R2 bucket for re-listen, or delete it after transcription. Either way, your transcript stays accessible.
MP3, MP4, MOV, WAV, M4A, FLAC, WebM, MKV, OGG and more. Up to 5GB and 10h per file. Auto-detects format and metadata.
You may also like
300 free minutes per month. No credit card. Iberian + LATAM Spanish ready.
Try for free