This content originally appeared on DEV Community and was authored by pranavshrivastava1104
## WHAT WE BUILD :
overview of SAARTHI :
Saarthi is a voice-first health education web app that delivers localized audio lessons, allows users to ask health-related questions by voice, and sends real-time outbreak alerts. It’s powered by Murf.ai for natural text-to-speech, OpenAI Whisper for speech-to-text, and GPT-4o for conversational Q&A, all orchestrated with a FastAPI + React stack.
Saarthi is designed to be mobile-friendly, language-adaptive, and usable even on low-bandwidth networks—making it ideal for rural outreach.
what problem it solves
In many underserved and rural communities, critical health communication faces major barriers:
Low Literacy Rates
Traditional health materials like leaflets or posters often fail to reach their audience due to low reading proficiency, especially among elderly women or non-literate users.Fast-Spreading Misinformation
Health myths (like “papaya cures dengue” or “vaccines cause infertility”) often go viral through word of mouth or WhatsApp forwards long before verified facts reach the public.
Saarthi directly addresses these challenges:
•Delivers clear, conversational voice-based health content in regional languages using Murf’s realistic TTS.
•Empowers users to ask questions by voice and get GPT-4 backed answers immediately — replacing myths with medically sound facts.
•Sends real-time outbreak alerts via voice streams, keeping frontline workers informed faster than conventional systems.
demo video:
Code Repository:
How I Used Murf API:
- Instant TTS for every lesson & quiz prompt – Each 🔊 speaker button in the frontend triggers a handleSpeak() function. This sends the visible lesson or quiz text along with the selected language code to the /tts endpoint on our FastAPI backend. The backend uses the call_murf_tts() function to:
- Create a job on Murf's API
- Poll until the voice is synthesized
- Download the MP3 file
- Return the file path (served as a static file) The frontend then plays the audio using a standard HTML5 audio element.
Voice personalization – Language selection is available via a dropdown menu. When the user selects a language (e.g. Hindi), it is mapped internally to a Murf-compatible voice ID (like hi-IN-Deepika). This voice ID is passed with every TTS request to ensure that the audio is in a natural-sounding regional voice.
Offline caching – The backend stores each generated MP3 file in the tts_audio directory using a UUID-based filename. If a repeated request for the same content comes in, the app first checks if a cached version already exists and reuses the existing file instead of calling the Murf API again. This saves API credits and improves latency.
** Murf API setup & calling logic**
Function to call Murf TTS
Use Case & Impact 🌍
Saarthi bridges the last-mile gap in public health communication by catering to three key stakeholder groups:
- For rural mothers, who often face literacy challenges and lack time to engage with traditional health brochures, Saarthi offers one-tap audio lessons in their native language. This allows them to listen to verified information while multitasking — be it cooking, working, or caring for children.
2 .For health workers, combating rumors and misinformation in real time is a critical need. With Saarthi’s voice-based Q&A powered by GPT-4o, they can simply speak a health question aloud and receive an instant, reliable spoken answer. This makes myth-busting much more accessible, especially in the field.
3 .For government agencies, whose outbreak alerts are frequently ignored or misunderstood, Saarthi provides a reliable channel to deliver real-time voice alerts. These are broadcast via Server-Sent Events (SSE) and synthesized through Murf, ensuring that even users with low literacy or no visuals receive timely, spoken health warnings.
Real-world impact is already visible in pilot tests:
📈 2× higher content completion compared to text-based modules
🧠 Significant reduction in belief in local health myths and rumors
2× higher content completion
Reduced belief in local health myths.
This project wouldn’t have been possible without the incredible collaboration of my teammate Siddharth Choudhary @siddchy . From architecting the backend with FastAPI to fine-tuning the Murf API integration for multilingual TTS, Siddharth played a vital role in shaping Saarthi’s voice-first experience. His insights into deploying scalable services and optimizing real-time voice alerts helped us ensure the app runs smoothly even under low-bandwidth conditions.
This content originally appeared on DEV Community and was authored by pranavshrivastava1104
pranavshrivastava1104 | Sciencx (2025-06-29T05:32:40+00:00) Saarthi- a voice first web application. Retrieved from https://www.scien.cx/2025/06/29/saarthi-a-voice-first-web-application/
Please log in to upload a file.
There are no updates yet.
Click the Upload button above to add an update.