The "I speak Language X" card represented New York City's primary strategy for assisting residents who speak relatively uncommon languages (i.e., languages other than English, Spanish, French, or German). Although it was the most concerted effort to reach these speakers, the initiative was undermined by the limitations of written communication -- low literacy, inconsistent orthographies in those languages, and other structural constraints that impeded accurate identification. I was intrigued by the recent development in the TTS (text-to-speech) technology and I wondered if I could make the audio version of "I speak Language X."
This page presents multilingual audio prompts that ask whether a listener speaks a given language. Each prompt was generated with both OpenAI text-to-speech and ElevenLabs. About 40 languages were chosen based on frequent non-English languages.
English source prompt
We are here to help you in the language you feel most comfortable using. Do you speak Language X? If yes, please say "yes." If not, please say "no."
Languages39
OpenAI files39
ElevenLabs files39
Reviewed translations1
High-stakes use warning
These files are for research, demonstration, and language-preference identification. They are not a substitute for qualified interpreters. The translations are draft machine translations and should be reviewed by qualified speakers before use in hospitals, schools, police interactions, or other high-stakes settings.
OpenAI developed Whisper, an open-source transcription program, to train its large language model (i.e., ChatGPT). OpenAI needed Whisper because they exhausted all the text data for training ChatGPT and decided to use audio data for training. After the launch of the web-interface for ChatGPT, OpenAI developed the tts (text-to-speech) models using the same architecture as Whisper. The tts-1 and tts-1-hd models were released in 2023, followed by gpt-4o-mini-tts in 2025. There are 13 different voices in gpt-4o-mini-tts. Unlike ElevenLabs, none of them have been trained for any specific language (except English, which is their primary language).
ElevenLabs is an AI research company known for its strong text-audio models such as TTS, voice cloning, and speech transcription. The latest model, v3, was released in 2026 and features natural-sounding English and a wider coverage of languages other than English. The signature feature of ElevenLabs is "voices," which can be trained for specific purposes, such as cloning your own voice or speaking in a specific language or accent.
Generated from the translated prompt using per-language voice choices in the CSV:
আপনি যে ভাষায় সবচেয়ে স্বচ্ছন্দ বোধ করেন, সেই ভাষায় আপনাকে সাহায্য করার জন্য আমরা এখানে আছি। আপনি কি বাংলা বলতে পারেন? যদি পারেন, অনুগ্রহ করে "হ্যাঁ" বলুন। যদি না পারেন, অনুগ্রহ করে "না" বলুন।
तपाईंलाई सबैभन्दा सहज लाग्ने भाषामा सहयोग गर्न हामी यहाँ छौँ। के तपाईं नेपाली बोल्नुहुन्छ? बोल्नुहुन्छ भने कृपया "हो" भन्नुहोस्। बोल्नुहुन्न भने कृपया "होइन" भन्नुहोस्।
نحن هنا لمساعدتك باللغة التي تشعر براحة أكبر في استخدامها. هل تتحدث العربية؟ إذا كان الجواب نعم، من فضلك قل "نعم". وإذا لم تكن تتحدثها، من فضلك قل "لا".
Narito kami para tulungan ka sa wikang pinakakomportable mong gamitin. Nagsasalita ka ba ng Filipino o Tagalog? Kung oo, pakisabi ang "oo". Kung hindi, pakisabi ang "hindi".
Jesteśmy tutaj, aby pomóc Państwu w języku, w którym czują się Państwo najbardziej komfortowo. Czy mówi Pan/Pani po polsku? Jeśli tak, proszę powiedzieć "tak". Jeśli nie, proszę powiedzieć "nie".
Nous sommes ici pour vous aider dans la langue dans laquelle vous vous sentez le plus à l'aise. Parlez-vous français ? Si oui, veuillez dire "oui". Sinon, veuillez dire "non".
Мы здесь, чтобы помочь вам на языке, на котором вам удобнее всего общаться. Вы говорите по-русски? Если да, пожалуйста, скажите «да». Если нет, пожалуйста, скажите «нет».
ہم یہاں آپ کی مدد اس زبان میں کرنے کے لیے موجود ہیں جس میں آپ سب سے زیادہ آرام محسوس کرتے ہیں۔ کیا آپ اردو بولتے ہیں؟ اگر ہاں، تو براہِ کرم "ہاں" کہیں۔ اگر نہیں، تو براہِ کرم "نہیں" کہیں۔
Estamos aqui para ajudar você no idioma em que se sente mais confortável. Você fala português? Se sim, por favor diga "sim". Se não, por favor diga "não".
Είμαστε εδώ για να σας βοηθήσουμε στη γλώσσα στην οποία αισθάνεστε πιο άνετα. Μιλάτε ελληνικά; Αν ναι, παρακαλώ πείτε «ναι». Αν όχι, παρακαλώ πείτε «όχι».
हम आपकी मदद उस भाषा में करने के लिए यहाँ हैं जिसमें आप सबसे अधिक सहज महसूस करते हैं। क्या आप हिंदी बोलते हैं? अगर हाँ, तो कृपया "हाँ" कहें। अगर नहीं, तो कृपया "नहीं" कहें।
Suntem aici să vă ajutăm în limba în care vă simțiți cel mai confortabil. Vorbiți română? Dacă da, vă rugăm să spuneți "da". Dacă nu, vă rugăm să spuneți "nu".
Ovdje smo da vam pomognemo na jeziku na kojem se osjećate najugodnije. Govorite li srpskohrvatski? Ako da, molimo recite "da". Ako ne, molimo recite "ne".
Consider separate Serbian/Croatian/Bosnian audio targets if the project expands.
#24 · uk
Ukrainian (українська)
draft_machine_translationunreviewed
Translated prompt
Ми тут, щоб допомогти вам тією мовою, якою вам найзручніше користуватися. Ви говорите українською? Якщо так, будь ласка, скажіть «так». Якщо ні, будь ласка, скажіть «ні».
Biz sizga o'zingizni eng qulay his qiladigan tilda yordam berish uchun shu yerdamiz. O'zbek tilida gapirasizmi? Agar ha bo'lsa, iltimos "ha" deng. Agar yo'q bo'lsa, iltimos "yo'q" deng.
Size en rahat hissettiğiniz dilde yardımcı olmak için buradayız. Türkçe konuşuyor musunuz? Eğer konuşuyorsanız lütfen "evet" deyin. Konuşmuyorsanız lütfen "hayır" deyin.
અમે તમને તે ભાષામાં મદદ કરવા અહીં છીએ જેમાં તમને સૌથી વધુ આરામદાયક લાગે છે. શું તમે ગુજરાતી બોલો છો? જો હા, તો કૃપા કરીને "હા" કહો. જો નહીં, તો કૃપા કરીને "ના" કહો.
A wa nibi lati ran yin lowo ni ede ti inu yin dun julo lati lo. Se e n so ede Yoruba? Ti bee ni, e jowo e so pe "bee ni". Ti bee ko ba ri, e jowo e so pe "rara".
Kami di sini untuk membantu Anda dalam bahasa yang paling nyaman Anda gunakan. Apakah Anda berbicara bahasa Indonesia? Jika ya, silakan katakan "ya". Jika tidak, silakan katakan "tidak".
LaGuardia label is Wu; confirm public wording before publication.
#32 · nan
Min Chinese / Fujian dialect (閩南語 / 福建話)
draft_machine_translationunreviewed
Translated prompt
Goan ti chia beh pang-cho li eng li siang kam-kak khah chhin-chhai e gi-gi. Li e hiau kong Ban-lam-gi, Hok-kian e gi-gi bo? Na e hiau, chhiann kong "u". Na bo hiau, chhiann kong "bo".
Ngai-den hai lia-e, yung ngi zhung suk-sit e ngien-ngi loi bong-zhu ngi. Ngi oi gong Hak-ka-fa mo? Oi e va, chhiang gong "oi". M oi e va, chhiang gong "m oi".
உங்களுக்கு மிகவும் வசதியாக இருக்கும் மொழியில் உதவுவதற்காக நாங்கள் இங்கே இருக்கிறோம். நீங்கள் தமிழ் பேசுகிறீர்களா? பேசினால், தயவுசெய்து "ஆம்" என்று சொல்லுங்கள். பேசவில்லை என்றால், தயவுசெய்து "இல்லை" என்று சொல்லுங்கள்.
നിങ്ങൾക്ക് ഏറ്റവും സൗകര്യമായി തോന്നുന്ന ഭാഷയിൽ നിങ്ങളെ സഹായിക്കാൻ ഞങ്ങൾ ഇവിടെ ഉണ്ട്. നിങ്ങൾ മലയാളം സംസാരിക്കുമോ? സംസാരിക്കുന്നുവെങ്കിൽ ദയവായി "അതെ" എന്ന് പറയുക. ഇല്ലെങ്കിൽ ദയവായി "ഇല്ല" എന്ന് പറയുക.
Adda kami ditoy tapno tulongan ka iti pagsasao a kasayaatan ken kasangayan mo nga usaren. Agsasaoka kadi iti Ilocano? No wen, pangngaasim ta kunaem ti "wen". No saan, pangngaasim ta kunaem ti "saan".
Chúng tôi ở đây để giúp quý vị bằng ngôn ngữ mà quý vị cảm thấy thoải mái nhất khi sử dụng. Quý vị có nói tiếng Việt không? Nếu có, xin hãy nói "có". Nếu không, xin hãy nói "không".