Contact person: Business Client

Phone:Show

Email:Show

Location: Tiruvannamalai, India

Budget: Recommended by industry experts

Time to start: As soon as possible

Project description:
"We are developing a prototype of an Indian language speech recognition and translation system using open-source technologies only (no Google, AWS, or Azure APIs).

The system should be capable of:

Converting speech to text in multiple Indian languages

Translating text between selected Indian languages

Providing basic speaker identification (diarization)

Offering a RESTful API interface for integration


This is a proof-of-concept (MVP) project focusing on 5–10 major Indian languages such as Hindi, Tamil, Telugu, Bengali, and Marathi. The goal is to build a functional base system that can later be expanded to cover 150+ languages and dialects.


---

Scope of Work

1. Audio Processing

Handle input formats (MP3, WAV, FLAC, M4A)

Perform noise reduction and normalization to 16kHz

Prepare data for speech recognition models



2. Speech Recognition (ASR)

Implement speech-to-text using open-source models (Meta MMS, Whisper, or Coqui STT)

Support multiple Indian languages

Provide language detection and transcription confidence scores



3. Text Translation

Use open-source translation models such as IndicTrans2 or MarianNMT

Enable bidirectional text translation between selected Indian languages



4. Speaker Diarization

Integrate speaker detection using [login to view URL] or a similar open-source tool



5. API Development

Develop RESTful API endpoints for speech-to-text and translation

Include basic authentication and documentation



6. Deliverables

Complete source code (Python preferred)

Deployment and configuration scripts

Technical and API documentation





---

Preferred Tech Stack

Python, FastAPI, PyTorch

Whisper, MMS, IndicTrans2, [login to view URL]

Hugging Face Transformers

Docker for deployment



---

Deliverables and Timeline

Functional MVP covering 5–10 Indian languages

Duration: 6–8 weeks

Budget: ₹40,000 (fixed)



---

Required Skills

Experience with speech recognition and translation using open-source models

Familiarity with Indian language datasets (Bhashini, AI4Bharat, Common Voice)

Strong Python and API development experience

Ability to deliver clean, documented, and reproducible code



---

How to Apply

Please include the following in your proposal:

1. Short summary of relevant experience with ASR or translation systems


2. Example projects or GitHub repositories


3. Proposed timeline and approach for the MVP" (client-provided description)