Speech-to-Text
Accurate transcription built for Indian accents, dialects, and real-world audio with speaker diarisation.
Text-to-Speech
Natural, expressive voices in 10 Indian languages and English, for conversational agents, dubbing, and accessibility.
Language model
Multilingual model with support for Indian languages, for text generation and reasoning.
Translation
High-quality translation across 12+ Indian languages, optimised for conversational and enterprise content.
Transcribe audio in 23 Indian languages
Upload audio from a form, pull it from a webhook, or fetch it from storage. Sarvam's speech to text handles Hindi, Tamil, Bengali, Telugu, and 19 more, with modes for standard transcription, code-mixed output, Roman transliteration, or direct translation to English for downstream tools that only read one language.
Synthesize speech that sounds right for Indian users
Turn any text into audio across 11 languages and dozens of speakers. Set the pace, pick the codec, and choose a voice that fits your product. The output goes directly into a Respond to Webhook node, an email, or a storage bucket, wherever your workflow sends audio next.
Run Sarvam's language models on any text
Feed transcripts, form inputs, or scraped content into sarvam-105b for high-quality multilingual reasoning. Chain completions with speech to text and text to speech to build voice Q&A flows, automated call handling, or multilingual support automations, all in the node editor.
Most Sarvam automations follow the same shape.
Receive audio
A webhook, form, or file download brings audio into the workflow as a binary property.
Transcribe
Sarvam speech to text reads that binary and returns a transcript.
Process
The chat node takes the transcript as a user message and returns a reply.
Respond
Text to speech reads the reply and returns audio you can send, store, or stream back.
No custom backend. No glue code. The node editor wires it.
Transcribe a form audio upload
A Form Trigger captures the file. Sarvam transcribes it and returns text in JSON. Route it to a CRM, a support ticket, or a Google Sheet.
Webhook voice Q&A
An incoming POST with raw audio goes to Sarvam speech to text, then chat, then text to speech. Respond to Webhook returns the spoken answer. Full round-trip in four nodes.
Batch transcription from storage
Pull audio files from S3 or Google Drive. Loop through with Sarvam speech to text. Write transcripts to a database or sheet. Useful for call centres, field recordings, or interview archives.
Multilingual support automation
Sarvam's Translate to English mode converts a call in any supported Indian language to English text on transcription. Feed it straight to your existing English-only model or ticketing system. No language detection step, no translation middleware.
How to install
The Sarvam node is available as a community package on n8n.
Open community nodes
Go to Settings, then Community nodes.
Install the package
Search for n8n-nodes-sarvam and install it.
Restart n8n
Restart n8n if prompted.
Add credentials
Go to Credentials, create a new Sarvam AI API credential, and paste your API key from indus.sarvam.ai.
Start building
Add a Sarvam node to any workflow and pick your Resource and Operation.
{
"resource": "speech",
"operation": "speechToText",
"binaryPropertyName": "audio",
"sttMode": "transcribe",
"speechToTextOptions": {}
}Match binary field names exactly
Input Binary Field on speech to text must equal the binary field name from the upstream node, such as data or audio. Check the Binary tab instead of guessing.
Name your Sarvam nodes
Rename each node to something like Sarvam STT or Sarvam Chat so $('node name') references survive copy and paste.
Pin sample data while you build
Pin a small audio file on the trigger so you can tune speech to text options without uploading on every run.
Keep keys in credentials
Store the API key in the Sarvam AI API credential, never in expressions or exported workflow JSON. Use n8n external secrets in production.
Call
Call transcription pipelines
Transcribe recorded calls and route the transcripts to your CRM, a ticket, or a sheet.
Multilingual
Multilingual support automation
Use translate to English mode so English-only tools can read calls made in any supported Indian language.
Voice
Voice question and answer flows
Chain speech to text, chat, and text to speech behind a webhook and return a spoken answer.
Ready to automate? Follow the n8n guide in the Sarvam docs.
Ready to automate?
Follow the n8n guide in the Sarvam docs.