Keep this tab open — you'll be redirected to your transcript.
Latin AmericanCastilianRioplatenseCaribbean
Dialect coverage: both sides of the Atlantic
Spanish is one of the model's strongest languages — second only to English in training coverage. Mexican, Colombian, Rioplatense (Argentina/Uruguay, including voseo), Caribbean, Andean, and Castilian Spanish all transcribe at the model's top accuracy tier. Fast Caribbean speech with dropped syllable-final s is the hardest case; expect a few more fixes there.
Distinción vs seseo doesn't matter to the output: the transcript is written in standard orthography either way, with ñ, accent marks, and inverted ¿¡ placed by the model.
Spanglish and code-switching
Real conversations flip between Spanish and English mid-sentence. With the language pinned to Spanish (as it is on this page), English fragments are usually transcribed as English words inside Spanish text rather than being mangled — but long English passages come out better if you use the auto-detect transcriber on the home page instead, which switches per file.
Frequently asked questions
Does it handle Latin American and Castilian Spanish equally?
Yes. The training data skews slightly Latin American by volume, but Castilian (including distinción) transcribes at the same tier. Regional vocabulary — vosotros forms, voseo conjugations, Mexican and Argentine slang — is written as spoken.
How accurate is Spanish transcription really?
Spanish sits in the model's top accuracy tier, typically within a point or two of English on clear audio — near-human for broadcast-quality speech. Accuracy dips on very fast Caribbean speech and heavy crosstalk, which is what the built-in editor is for.
Will accents and ñ come out right in the text?
Yes. Output is standard Spanish orthography: tildes (á, é, í, ó, ú), ñ, ü, and inverted question/exclamation marks are all produced by the model automatically. You get properly written Spanish, not a stripped ASCII approximation.
What about Spanglish — mixed Spanish and English?
Short English phrases inside Spanish speech ("okay", brand names, work jargon) are transcribed inline correctly on this page. If your recording is genuinely half-and-half, use the auto-detect uploader on the home page — pinning Spanish forces borderline sentences toward Spanish spellings.
Can I transcribe Spanish audio directly into English text?
Not on this page — this transcribes Spanish speech into Spanish text, which is the accurate path. Machine-translating the transcript afterwards (with any translator you like) preserves far more meaning than asking a speech model to translate on the fly. Transcript translation is on our roadmap.
Is Catalan, Galician, or Basque covered here?
Not by this page — they're separate languages, not Spanish dialects. The model does support Catalan, Galician, and Basque: upload on the home page with auto-detect, or pin the language in the options there.