&imwidth=800&imheight=600&format=webp&quality=medium "Sarvam AI launches Saaras V4: 5 features that set the new AI model apart")
India's voice AI race has a new contender. Sarvam AI has launched Saaras V4, its latest automatic speech recognition (ASR) model, designed to transcribe speech across Indian languages, English, regional accents and noisy environments.
The model can produce transcripts in five formats, including verbatim text, translation and code-mixed language output. Sarvam says it can also identify spoken languages automatically and process audio in real time, with streaming latency below 150 milliseconds. The launch targets a major challenge for voice technology in India: understanding people who switch between languages, speak with regional accents or use everyday expressions that traditional transcription systems can struggle with.
Sarvam Saaras V4 brings five speech output modes
Saaras V4 can convert the same audio into five different formats: verbatim transcription, normalised text, code-mixed text, transliteration and translation. This means developers can choose whether to preserve exactly what someone says, clean up the text, write spoken words in another script or translate them into another language. Sarvam says these functions are built into the model, rather than requiring separate systems to process the transcript afterwards. The model combines an audio encoder with a 3-billion-parameter hybrid state-space language model. It is designed to handle code-switching, dialect differences and background noise.
22 Indian languages and a focus on accuracy
Saaras V4 can automatically identify a spoken language and transcribe it in the corresponding native script. On verified IndicVoices data, Sarvam reports a language identification error rate of 5.22% across 22 Indian languages, falling to 2.9% across the 10 most widely spoken languages in its evaluation. For English, the company says Saaras V4 recorded the lowest average word error rate across seven benchmarks covering Indian English, international accents, meetings, financial conversations and media. The model was also tested on the Vistaar benchmark across 10 Indian languages. These are company-reported results, and performance in everyday use may vary depending on accents, recording quality and background noise.
Trending Stories
)
Outrage erupts after 3-year-old is married to 60-year-old man in Pakistan, link to her sister's murder being probed
)
Trump-Xi summit wrap: Big pageantry, Iran talks, few major developments and more meetings planned
)
Did Hitler kill his niece Geli Raubal? Newly found records reopen a 95-year-old mystery
)
‘Attack on Mecca a red line for Pakistan’: Shehbaz Sharif warns Houthis at UN
)
Nigeria Navy moves to boost oil security, tackle crude theft and infrastructure threats
)
US Supreme Court backs Trump administration's use of SAVE database to check voter citizenship
)
'A terrorist country is asking me that question?': Jaishankar hits back at Pakistani journalist at UN; flags attacks on commercial shipping
)
Cindy Crawford’s son Presley Gerber cremated as death probe continues
)
US court backs Pentagon's decision to exclude Anthropic over Claude AI restrictions
)
'Making history': Xi Jinping praises Trump as US-China leaders dine at White House
)
India's UN Envoy slams Pakistan: 'They claim security by day, terrorism by night'
)
Saudi Arabia still harbours intent to develop nuclear weapons, reveals US intelligence: Report
)
India flags brutal violence in Pak-occupied Jammu Kashmir
)
WATCH | As Shehbaz Sharif claims Pakistan’s ‘victory’ during 2025 conflict, Indian diplomat R Srinivas laughs at UN
)
Red Sea crisis: Houthis promise not to target European ships; UNSC condemns Saudi attacks
)
Yemen's war escalates: Saudi-backed leader calls for general mobilisation against the Houthis
)
Trump rejects Iran's 7-day ceasefire proposal as US weighs renewed bombing after midterms: Report
)
Asian Games 2026: India women’s kabaddi team edge Iran 37-34 to clinch 3rd gold medal
)
Gaza's war turned environmental damage into second crisis
)
‘Not just a local tragedy’: Nepal PM Shah warns UN of growing climate threat, says Nepal floods were 'a warning to the world'
)
'FIRST LOOK from inside Noah’s Ark': Researchers share video from inside mysterious mound in Turkey
)
Jay-Z rape case: Accuser says allegations were false, never met rapper
)
Heart of the Beast review: Brad Pitt leads a gripping tale of survival and loyalty
)
India's insurance regulator overhauls policies
)
Myanmar training camp case: US national, 6 Ukrainians allowed to meet their families abroad
)
‘Goal is to win it’: Virat Kohli confirms 2027 ODI World Cup will be his final tournament for India
)
‘It is all possible’: Putin weighs next steps on Ukraine talks, rules out Europe conflict
)
Asian Games: Shooting spree continues as India win silver in women’s 50m rifle 3P team event
)
Saudi oil exports surge despite Iran war
)
Scientists detect first radio signals directly from faraway planet
&im=FitAndFill=(700,400))
New Friday OTT releases (September 25, 2026): Don’t Be Shy, Hunkkaar: The Roar, Mango Pachcha- 6 movies and TV shows on Netflix, Prime Video and more
)
10,000 gather in support of 87-year-old forcefully evicted from home of 71 years
)
Pakistan deploys over 20,000 security personnel as PTI vows march for Imran Khan's release
)
Pentagon quietly adds 37 wounded troops to Iran war tally without explaining when or how
)
Indian diplomat Petal Gahlot raises deteriorating situation in Pakistan-occupied Kashmir in right of reply at UN
)
Mushtaq Khan dies at 56: When Welcome actor opened up about his salary for iconic role Ballu
)
Tesla is preparing something big for Optimus: new Gen 3 models surface in app
)
OpenAI AI Agents hacked 4 websites before Hugging Face, researchers reveal
Faster voice AI for developers
Sarvam says Saaras V4 supports streaming with time to first token below 150 milliseconds, alongside long-form audio processing. It also includes keyterm prompting, allowing developers to supply names, product terms and acronyms that may otherwise be difficult for speech recognition systems to identify. Developers can access the model through Sarvam AI's API, with Python and Node.js SDKs. Integrations include Vercel AI SDK, LiveKit Agents and Pipecat Agents. The launch could support applications such as voice assistants, multilingual customer service, meeting transcription and real-time translation. However, the announcement does not establish pricing, independent benchmark results or how the model compares with competing systems in real-world deployments.
