Multilingual AI Voice Rewriter

I am kicking off Phase 1 of a larger project and need a solid, production-ready module that can do three things seamlessly:

• Detect and understand spoken Hindi and English.
• Convert that speech to clean, accurate text in real time.
• Pass the raw transcript through an AI layer that can rephrase or correct grammar on demand before returning the final output.

The flow should feel natural to the user: speak, see text appear, tap (or say) “rephrase,” then receive polished copy instantly. I’m open to whichever stack you feel offers the best latency and accuracy—Google Speech-to-Text, Azure, Whisper, or your proven alternative—so long as it supports the two languages above and can run at scale. For the rephrasing engine, GPT-class quality is the baseline I’m aiming for.

Please outline:
1. Your proposed tech stack and why it fits.
2. Estimated turnaround time to deliver a testable MVP plus the timeline to move it to production quality.
3. A clear cost breakdown by milestone.

Phase 1 is my immediate priority. If we work well together you’ll have first shot at the e-commerce, payments, and crypto modules that follow. I’m ready to start as soon as I review your plan.