Live Hinglish Transcription Website
Budget: ₹1,500 – ₹12,500 INR
I’m looking for a custom-built web app that turns any spoken Hinglish—whether it comes from my microphone or the audio playing on my computer—into on-screen text instantly. The transcription must keep Hindi words in Devanagari while leaving English words in the Roman script, exactly like this sample snippet:
noida के कुछ ऐसे sectors में बाढ़ आई है कि जिनके करोड़ों के घर हैं, वो उन करोड़ों के घर के सामने इतना-इतना पानी भरा हुआ है…
Scope
• Real-time transcription is the only feature I need right now; no additional languages, history logs, or text editors.
• The page should start listening as soon as I hit “Start” and stream the output line by line with minimal latency.
• Audio can come from system sound or mic, so please integrate a solution comparable to screen-share audio capture (Web Audio API, WebRTC, or a lightweight helper where browsers restrict direct system capture).
• Accuracy in code-mixed contexts is key; you’re free to combine APIs or models (e.g., Google Speech-to-Text, Azure, Whisper) as long as the final text respects the Hindi/English split.
• The UI can stay simple—just a clean panel that shows the live feed and a “Copy All” button.
Deliverables
1. Front-end (HTML/CSS/JS) with the live transcript window.
2. Back-end service or worker that streams recognition results.
3. Deployment guide so I can run it on my own server (Linux preferred).
4. Brief README outlining any API keys, rate limits, or model setup.
I’ll test by playing a five-minute Hinglish news clip; acceptance hinges on seeing the mixed-script output appear in real time with a lag under two seconds.
noida के कुछ ऐसे sectors में बाढ़ आई है कि जिनके करोड़ों के घर हैं, वो उन करोड़ों के घर के सामने इतना-इतना पानी भरा हुआ है…
Scope
• Real-time transcription is the only feature I need right now; no additional languages, history logs, or text editors.
• The page should start listening as soon as I hit “Start” and stream the output line by line with minimal latency.
• Audio can come from system sound or mic, so please integrate a solution comparable to screen-share audio capture (Web Audio API, WebRTC, or a lightweight helper where browsers restrict direct system capture).
• Accuracy in code-mixed contexts is key; you’re free to combine APIs or models (e.g., Google Speech-to-Text, Azure, Whisper) as long as the final text respects the Hindi/English split.
• The UI can stay simple—just a clean panel that shows the live feed and a “Copy All” button.
Deliverables
1. Front-end (HTML/CSS/JS) with the live transcript window.
2. Back-end service or worker that streams recognition results.
3. Deployment guide so I can run it on my own server (Linux preferred).
4. Brief README outlining any API keys, rate limits, or model setup.
I’ll test by playing a five-minute Hinglish news clip; acceptance hinges on seeing the mixed-script output appear in real time with a lag under two seconds.
Related categories:
JavaScript
CSS
Transcription
HTML
Hindi Translator
English (US) Translator
Web Development
WebRTC