Hinglish text to speech that actually sounds Hinglish

A short code-mixed passage typed in Roman script.

• Roman-script Indian-language text and English words mixed into an Indian-language script are normalised before generation. Without that step, models trained only on native-script audio tend to skip Latin-script words silently rather than pronounce them.
• Because Hinglish routes through the Indian-language model, its reference clips share Hindi's 12-second cap.
• Record your reference clip in Hinglish if you intend to generate Hinglish. A reference clip in a different language measurably degrades output quality, so a clean English clip is a worse starting point for Hinglish than a slightly noisier Hinglish one.

These are the exact prompts Mayantar shows you in the recorder. Each one is timed to land inside the 12-second limit at a natural pace, and each covers something cloning models need: relaxed conversational prosody, spoken digits and dates, and a question with rising intonation. Read one aloud in a quiet room and you have a usable reference clip.

1. Hello dosto, yeh meri awaaz hai aur main aaj ek chhota sa sample record kar raha hoon. Mujhe music sunna pasand hai.
2. Chalo thode numbers bolte hain: 1, 2, 3, 4 aur 5. Meri meeting subah 9:30 baje, 15 March ko hai.
3. Kya aapne kabhi socha hai ki ek simple recording kisi ki awaaz ko kaise capture kar leti hai? Mujhe yeh kaafi interesting lagta hai.

More detail on what makes a clip pass or fail is in our guide to recording a reference clip.