Storm is designed to work the way your team actually talks. Workers can speak in their own language and Storm handles the translation — no setup, no language switching, no extra steps.
This is particularly useful for multilingual teams, international projects, and workforces where English is not the primary working language.
How Multilingual Processing Works
Storm detects the spoken language automatically and translates content into the destination template language during processing.
Input — workers speak in any supported language
Output — Storm generates responses in the template's working language by default
No setup required — there is no language toggle or user-level language setting
Example: A worker speaks in Spanish. The form template is in English. Storm transcribes the Spanish audio, translates the content, and fills the form fields in English.
Overriding the Default Translation
To keep output in the original spoken language, state this explicitly during your recording:
"Write this in Spanish"
"Don't translate"
"Keep this in Portuguese"
Storm will follow the instruction and produce output in the requested language.
Supported Languages
Storm supports approximately 50 languages for voice input. Storm auto-detects the language of each voice note — there is no language selection required.
Language Family | Languages |
Romance | English (en), Spanish (es, es-419), French (fr, fr-CA), Portuguese (pt, pt-BR, pt-PT), Italian (it), Romanian (ro), Catalan (ca) |
Germanic | German (de), Swiss German (de-CH), Dutch (nl), Flemish (nl-BE), Swedish (sv), Danish (da), Norwegian (no) |
Slavic | Russian (ru), Ukrainian (uk), Polish (pl), Czech (cs), Slovak (sk), Bulgarian (bg), Belarusian (be), Bosnian (bs), Croatian (hr), Serbian (sr), Slovenian (sl), Macedonian (mk) |
Other European | Greek (el), Finnish (fi), Hungarian (hu), Lithuanian (lt), Latvian (lv), Estonian (et), Turkish (tr) |
East / Southeast Asian | Japanese (ja), Korean (ko), Mandarin Simplified (zh, zh-CN), Mandarin Traditional (zh-TW), Cantonese (zh-HK), Thai (th), Vietnamese (vi), Indonesian (id), Malay (ms), Tagalog (tl) |
South Asian | Hindi (hi), Bengali (bn), Marathi (mr), Tamil (ta), Telugu (te), Kannada (kn), Gujarati (gu), Urdu (ur) |
Middle Eastern | Arabic Modern Standard (ar), plus 17 regional variants including Gulf (ar-AE, ar-SA, ar-QA, ar-KW), Levantine (ar-SY, ar-LB), Egyptian (ar-EG), Maghrebi (ar-MA, ar-DZ, ar-TN), and more; Hebrew (he); Persian/Farsi (fa) |
The supported language list may expand over time as additional languages are added.
What Is Always Preserved Verbatim
Regardless of the template language or translation setting, Storm never translates the following:
Proper nouns — people's names, company names, supplier names (e.g. "Maria", "ABC Company", "Komatsu")
Equipment models and SKUs — e.g. "Makita XS-450", "Hilti TE 70-D AVR"
Asset IDs and serial numbers
Place names — e.g. "Brisbane", "Tokyo"
Standards and regulation codes — e.g. "AS 1170", "OSHA 1926.501", "ISO 14001"
Document type abbreviations — SWMS, ITP, RAMS, JSA stay as-is
Direct quotations — quoted content stays in its original language
Bilingual Field Labels
If a template uses bilingual field labels — for example "Description / Descripción" or "Hours / Horas" — Storm fills both sides and matches the separator used in the label. If the label uses a slash, the output uses a slash. If it uses parentheses or a newline, the output follows the same pattern.
Storm also matches the label's order — if the label reads "EN / JP", the content reads English first, then Japanese. If the user only provided content in one language, Storm fills that side and leaves the other blank rather than fabricating a translation.
Questions About How Storm Handles Different Languages
How accurate is Storm with accented speech?
Regional accents are handled well — Storm uses dedicated regional variants for major languages (e.g. Latin-American Spanish vs. Iberian Spanish, Brazilian vs. European Portuguese, Canadian vs. European French). Heavy or non-native accents can reduce accuracy on uncommon vocabulary.
For best results, record in a quiet environment and speak proper nouns clearly. If accuracy is consistently poor on a specific accent, report the example to Sitemate Support — these become test cases that improve the model.
Does Storm work with construction slang in other languages?
Yes, with some variation by language. English construction terms (SWMS, ITP, RAMS, JSA, trade equipment names) are handled well. Spanish, Portuguese, French, German, Italian, and Japanese trade vocabulary also performs well — trade verbs and adjectives translate cleanly.
For less common supported languages, standard vocabulary works reliably but very local site slang may be transcribed phonetically. Where Storm isn't confident on a trade term, it keeps the source-language word rather than guessing at a translation.
What if Storm translates something incorrectly?
All AI Form Fill generated content is fully editable before saving. If a translation looks wrong, correct the field manually before submitting the form.
The original recording transcript is also stored verbatim in the form activity feed — never translated, never edited — so you can always check exactly what was said. If a translation issue recurs, contact Sitemate Support with the audio and form combination.
Will bilingual field labels be filled in both languages?
Yes. Storm detects bilingual templates automatically and fills every bilingual field in both languages, side by side, using the separator and order from the label. A template can mix monolingual and bilingual fields and Storm honours each individually.
What happens if I switch languages mid-recording?
Storm reconciles the meaning across both language fragments into a single intent, then writes the field in the template's target language. For example, a recording that starts in English and switches to Spanish mid-sentence produces a single, clean English output. The activity feed preserves the original mixed-language recording verbatim for audit purposes.
Can I mix English and Spanish in the same voice note?
Yes — this is called code-switching and is fully supported. The same applies to any combination of supported languages. Storm writes the form output in the template's working language while preserving proper nouns, quoted content, and identifiers in their original language.
Will Storm translate everything or just some parts?
Storm translates form field content into the template's target language. It does not translate activity feed transcripts, proper nouns, equipment models and SKUs, place names, standards codes, document type abbreviations (SWMS, ITP, RAMS), or direct quotations. These are always preserved verbatim in their original language.
What happens to quotes or names in a different language?
Quotations are preserved verbatim in the source language even when the surrounding content is translated. For example, a Spanish recording that references an English sign reading "DANGER HIGH VOLTAGE" produces an English output that keeps the sign text in English. This applies to incident reports too — if a worker quotes something verbatim, Storm preserves the exact words.
Do I need to set up a language in my account?
No. Storm detects the spoken language automatically. There is no language toggle or user-level language setting.
Current Limitations
Storm outputs in the template's language by default. To change this for a specific recording, state it explicitly — for example, "write this in Spanish."
Individual table cells cannot currently be excluded from translation.
Storm extracts text from photos into form fields but does not translate text visible inside the photo itself.
Mongolian is not supported for voice input. Typed/written input is supported for Modern Mongolian script only — Traditional Mongolian (vertical script) is not supported.
