Today, we're introducing the Sarvam Creator Partner Programme, for creators who want their videos to reach audiences across languages.

Partners get free credits to dub their content in their own voice with Sarvam Content Studio, weekly briefs on what we're building, and rewards ng videos.

More about the programme: links.sarvam.io/g8e4mI

Sarvam x @figma

Agents built on Indus can now create designs directly in Figma, through MCP. They read your components, variables and layouts, then build with them.

Introducing Sarvam Voice Agents MCP.

Describe the voice agent you want, and your coding tool builds it. The MCP connects your coding tool to Sarvam, so you can create agents, run campaigns, and analyse calls from the environment you already code in.

Watch Sarvam Voice Agents , call a list of people and report back.

Build your first voice agent with the MCP: links.sarvam.io/2oxeo2

Crazy demos, big announcements, coolest merch, great coffee and even better company 💚

kivi - day 2 recap 🎥

Today, we're introducing Bulbul V4 Flash, our latest text to speech model built for natural, expressive speech across Indian languages.

Flash brings frontier-quality speech into a smaller, more efficient model, preserving the subtle details that make voices sound natural while n significantly faster.

See details in our blog: links.sarvam.io/siuhmS

A voice sounds native when it gets the small things right. Flash brings richer emotion, more natural pacing and context-aware delivery across Indian languages.

It handles regional accents, Indian English, names, numbers and mixed-language conversations, while delivering fast, consistent responses even across thousands of simultaneous calls.

We’re launching Bulbul V4 Flash at 50% off the standard price of ₹20 per 10,000 characters. Pay just ₹10 till 12 November, 2026.

Flash is available now through Sarvam APIs.

Try Bulbul V4 Flash → links.sarvam.io/s3AEny

You've been talking to kivi all this time.

Now see the people building it 😎

MEET THE KIVIES

thank you so much for giving kivi such a warm welcome 🎙️

now let's meet and talk more about it, this saturday, 10th october.

we've planned a lot for you, and we're not telling you what yet 👀

very limited spots → luma.com/a3bnkasg

come say hi. we're very ready to turn you into a kivi fan 💚

A good dub should feel like it was made in that language from the start.

In this video, we show how dubbing works in Content Studio, retaining each speaker’s original voice while translating the video into another language.

You can choose how conversational the translation ry line, and fix names, wording, speakers or timing before putting it out in the world. When you make a change, only that part regenerates, not the entire video.

Make your next video speak the language your audience prefers. Try Content Studio: links.sarvam.io/content-studio…

Dubbing is also available through the API.

Send a source video and a list of target languages to get back the dubbed video, audio track, and subtitles for each language.

Get started with the docs: docs.sarvam.ai/api/api-guides…

Kivi is now live on Mac.

Use your voice to get more done across your Mac.

Powered by Saaras V4, our latest Speech to Text model.

Try it at heykivi.ai

Sarvam Campus had a crazy start last weekend. 🚀

We kicked things off across IIT Madras, SRM IST, and NIT Trichy with hands-on workshops and build-along sessions.

3 campuses.
3 workshops.
450+ attendees.
200+ projects built in just 45 minutes.

Loved seeing what everyone built. started.

Introducing Saaras V4, our most capable speech to text model yet.

It delivers strong performance across both English and Indian languages, with better accuracy over a much wider range of speech.

Saaras V4 was built with specific attention to noise, accents, dialects and Read more in the blog: links.sarvam.io/bqvvyS

Saaras V4 preserves transcription accuracy across noisy audio.

We evaluated Saaras V4 on Kathbath Noisy using LLM-WER, across speech affected by compression, clipping and background interference.

Saaras V4 maintains substantially higher transcription accuracy under these or rate less than half that of Deepgram Nova-3 and GPT-4o Transcribe.

Saaras V4 also supports five transcript formats for different applications.

The formats let the same model serve very different workflows, from low-latency voice agents to verbatim transcripts for compliance and cleaner English for analysis.

Sarvam Vision 2.1 is live, with new capabilities for structured extraction and Indic handwritten recognition across English and 22 Indian languages.

The model is state of the art on both olmOCR-Bench and our own Indic benchmark, scoring 87.3 and 87.39 respectively.

Read more in the blog: links.sarvam.io/zF7SiK

Sarvam Vision 2.1 extracts key-value pairs directly from tables and forms, including multi-page tables and dense, multi-field layouts.

It also reads handwritten Indian-language text, transcribing it directly and pulling structured fields out of handwritten forms.