Pithflow's transcription model is multilingual-first, not English with bolted-on translations. 100+ languages including Spanish (LATAM + Spain), Portuguese (BR + PT), French, German, Japanese, Mandarin, Hindi, Arabic — all with auto-detection per clip.
Set the Languages tab to Auto-detect (the default). Every clip is analyzed for what language was actually spoken, and the transcription matches. Speak English at 9am, Spanish at 10am, code-switch mid-sentence at 11am — no settings to flip.
Spanish dictation gets inverted question marks (¿), inverted exclamations (¡), and correct accent placement (está vs esta). French gets non-breaking spaces before colons. German gets proper compound capitalization. The cleanup pass knows the rules per language, not just per code page.
Many bilingual conversations switch languages mid-sentence — "Sí, send me the report" or "Vamos a hacer un meeting tomorrow". Pithflow handles this gracefully because the transcription model sees the entire utterance, not just the first word's language.
If you dictated Spanish, the cleanup pass cleans up Spanish (Spanish filler words, Spanish tone presets). It doesn't silently translate to English. The model is explicitly instructed: preserve the language unless asked to translate.
Auto-detect occasionally misjudges very short clips ("Sí." or "OK") because there isn't enough audio to be confident. Lock the Languages tab to a specific language for sessions where you only want that language — Pithflow stops detecting and uses the locked language regardless.