🇮🇳हिन्दी

Indian TTS for everyone

High-quality Text-to-Speech for Indian languages using Sarvam AI.

⚠️ Required: API Key

You must have a Sarvam API key to use this skill.

Get your free API key at: https://dashboard.sarvam.ai

Setup

1. Go to Sarvam Dashboard and generate an API key
2. Add it to Zo's secrets: Settings → Advanced
3. Add secret: SARVAM_API_KEY = your-api-key

Document Intelligence

Extract text and structure from PDF documents and images (JPEG/PNG) using Sarvam AI's Document Intelligence API. Supports 23 Indian languages plus English.

Supported Languages

Output Formats

• md - Markdown files (default, human-readable)
• html - Structured HTML with layout preservation
• json - Structured JSON for programmatic processing

Usage

# Process a PDF document
python3 scripts/document_intelligence.py document.pdf --language hi-IN --format md

# Process with custom output directory
python3 scripts/document_intelligence.py document.pdf -o ./extracted/

# Check status of existing job
python3 scripts/document_intelligence.py --job-id <job-id>

# Download results for completed job
python3 scripts/document_intelligence.py --job-id <job-id> --download -o ./output/

File Constraints

• PDF files or ZIP files containing JPEG/PNG images
• Maximum file size: 200 MB
• Maximum pages/images: 500
• ZIP files must be flat (no nested folders)

Options

Text Processing

Text AI capabilities including chat completion, translation, transliteration, and language detection. Supports 23 Indian languages.

Chat / Text Completion

Sarvam's LLM APIs for chat and text completion. Two models available:

• sarvam-105b (Flagship) - Most capable model for complex reasoning, coding, and instruction following
• sarvam-m - Efficient model for general chat

sarvam-105b (Flagship Model)

The flagship 105B parameter model with state-of-the-art performance on Hindi and Indian language benchmarks.

# Chat using sarvam-105b (default)
python3 scripts/text_processing.py chat "Explain quantum computing in simple terms"

# Chat with system context
python3 scripts/text_processing.py chat "Write a poem" --system "You are a creative poet"

# Adjust temperature (0-2, lower = more focused)
python3 scripts/text_processing.py chat "Creative story" --temperature 0.8

Key Features :

• Best for complex reasoning, coding, and instruction following
• Excellent Hindi and Indian language understanding
• Supports 22 Indian languages + English
• Large context window

sarvam-m (Efficient Model)

Efficient model for general chat and text completion.

# Simple chat
python3 scripts/text_processing.py chat "What is the capital of India?" --model sarvam-m

# Chat with system context
python3 scripts/text_processing.py chat "Tell me about AI" --system "You are a helpful AI assistant" --model sarvam-m

# Adjust temperature
python3 scripts/text_processing.py chat "Creative story" --temperature 0.8 --model sarvam-m

Translation

Translate text between 23 Indian languages. Two models available:

• mayura:v1 - 12 languages, supports modes and transliteration
• sarvam-translate:v1 - All 23 languages, formal mode only

Languages : hi-IN, en-IN, bn-IN, gu-IN, kn-IN, ml-IN, mr-IN, od-IN, pa-IN, ta-IN, te-IN, as-IN, brx-IN, doi-IN, kok-IN, ks-IN, mai-IN, mni-IN, ne-IN, sa-IN, sat-IN, sd-IN, ur-IN

# Auto-detect source and translate to Hindi
python3 scripts/text_processing.py translate "Hello, how are you?" --target hi-IN

# Specify source language
python3 scripts/text_processing.py translate "नमस्ते" --source hi-IN --target en-IN

# Use Mayura model with colloquial mode
python3 scripts/text_processing.py translate "What's up?" --target hi-IN --model mayura:v1 --mode modern-colloquial

# Translate with romanized output
python3 scripts/text_processing.py translate "I am going home" --target hi-IN --model mayura:v1 --output-script roman

Modes (mayura:v1 only): formal, modern-colloquial, classic-colloquial, code-mixed

Output Scripts (mayura:v1 only): roman, fully-native, spoken-form-in-native

Transliteration

Convert text from one script to another while preserving pronunciation.

# Hindi to English (romanization)
python3 scripts/text_processing.py transliterate "नमस्ते" --source hi-IN --target en-IN

# English to Hindi
python3 scripts/text_processing.py transliterate "namaste" --source en-IN --target hi-IN

# With spoken form conversion
python3 scripts/text_processing.py transliterate "I have 2 meetings at 3pm" --source en-IN --target hi-IN --spoken-form

Language Detection

Automatically identify the language and script of text.

python3 scripts/text_processing.py detect "नमस्ते दुনিয়া"
# Output: Language: hi-IN, Script: Deva

python3 scripts/text_processing.py detect "Hello world"
# Output: Language: en-IN, Script: Latn

Speech-to-Text

Convert speech to text with automatic language detection and optional translation to English. Three modes available:

1. REST API - Quick transcription (<30 seconds), immediate results
2. WebSocket - Real-time streaming for live audio
3. Batch API - Process multiple files or longer audio with speaker diarization

Supported Languages (22 Indian languages)

hi-IN, bn-IN, kn-IN, ml-IN, mr-IN, od-IN, pa-IN, ta-IN, te-IN, gu-IN, en-IN, as-IN, ur-IN, ne-IN, kok-IN, ks-IN, sd-IN, sa-IN, sat-IN, mni-IN, brx-IN, mai-IN, doi-IN

REST API (Quick Transcription)

Best for short audio files (<30 seconds). Immediate results.

# Basic transcription with auto-translation to English
python3 scripts/speech_to_text.py rest audio.mp3

# With context prompt
python3 scripts/speech_to_text.py rest audio.mp3 --prompt "This is a conversation about technology"

# Specify codec for PCM files
python3 scripts/speech_to_text.py rest audio.raw --codec pcm_s16le

Supported formats : WAV, MP3, AAC, AIFF, OGG, OPUS, FLAC, MP4/M4A, AMR, WMA, WebM, PCM

WebSocket Streaming

Real-time speech-to-text with streaming audio.

# Streaming with translation (default)
python3 scripts/speech_to_text.py websocket audio.wav

# Transcription mode (no translation)
python3 scripts/speech_to_text.py websocket audio.wav --mode transcribe

# Different output modes (saaras:v3 only)
python3 scripts/speech_to_text.py websocket audio.wav --mode translit    # Romanized output
python3 scripts/speech_to_text.py websocket audio.wav --mode verbatim  # Exact word-for-word
python3 scripts/speech_to_text.py websocket audio.wav --mode codemix   # Code-mixed output

Modes (v3): translate, transcribe, verbatim, translit, codemix

Batch API (Long Audio & Multiple Files)

For longer audio or processing multiple files. Supports speaker diarization.

# Full workflow - create, upload, start, poll, download
python3 scripts/speech_to_text.py batch audio1.mp3 audio2.mp3 audio3.mp3 --output-dir ./transcripts/

# With speaker diarization
python3 scripts/speech_to_text.py batch meeting.wav --diarization --num-speakers 3

# Step-by-step workflow
# 1. Create job
python3 scripts/speech_to_text.py batch-create --diarization
# Returns: Job ID: abc-123

# 2. Upload files
python3 scripts/speech_to_text.py batch-upload abc-123 audio.mp3

# 3. Start job
python3 scripts/speech_to_text.py batch-start abc-123

# 4. Check status
python3 scripts/speech_to_text.py batch-status abc-123

# 5. Download results
python3 scripts/speech_to_text.py batch-download abc-123 output1.txt output2.txt --output-dir ./results/

Batch workflow states : Accepted → Pending → Running → Completed/Failed

Text-to-Speech (TTS)

High-quality Text-to-Speech for Indian languages using Sarvam AI's Bulbul v3 model.

Quick Start

python3 scripts/tts.py "नमस्ते, आप कैसे हैं?" --language hi-IN --speaker meera

Supported Languages

Speakers

Female : Meera (default), Priya, Neha, Simran, Kavya, Ishita, Shreya, Roopa, Tanya, Shruti, Suhani, Kavitha, Rupali, Amelia, Sophia

Male : Shubh, Aditya, Rahul, Amit, Dev, Arjun, Ratan, Varun, Manan, Sumit, Kabir, Aayan, Ashutosh, Advait, Anand, Tarun, Sunny, Mani, Gokul, Vijay, Mohit, Rehan, Soham

TTS Options

Resources

• Dashboard: https://dashboard.sarvam.ai
• Docs: https://docs.sarvam.ai
• Cookbook: https://github.com/sarvamai/sarvam-ai-cookbook

Weekly Installs

1

Repository

ankitjh4/indic-ai-skills

GitHub Stars

32

First Seen

1 day ago

Security Audits

Gen Agent Trust HubPassSocketFailSnykPass

Installed on

amp1

cline1

opencode1

cursor1

kimi-cli1

codex1