Reverie Unveils India‑Focused STT Model, 1.5× Faster Than Deepgram
India’s speech-to-text scene is a mess of scripts, Hindi-English code-switches and dozens of regional dialects that trip up most models. Reverie Language Technologies - a team that’s been training AI on Indian languages for over ten years - says its newest engine runs about 1.5 times faster than the well-known rival Deepgram. That could mean voice assistants that feel snappier, captioning that keeps up with streaming, and call-center tools that answer quicker, all while keeping the accuracy intact.
In the last year the service logged roughly 30 lakh API calls, hinting that developers and enterprises are already giving it a go. Still, the big question is whether it can handle the country’s endless accents and low-resource tongues at scale. The launch lines up with Reverie’s 16th anniversary, a neat milestone for a company still figuring out how far this tech can go.
The next paragraph spells out what the model actually does and why it matters.
Reverie Language Technologies, a veteran in Indian-language AI, marked its 16th anniversary with the launch of a new Speech-to-Text (STT) model built to decode India's multilingual chaos, from Hinglish to the several dialects. The model, which logged 30 lakh API calls over the last year, is tuned to India's real speech, full of code-switching, mixed languages, and regional quirks that global systems often fail to catch. In independent tests against Deepgram for voice agent use cases, Reverie's model scored about 4.2% higher in accuracy and 1.5x faster in response times, putting it among the most capable systems for Indian users.
Reverie’s new Speech-to-Text engine certainly sounds like it was built for India’s mix of Hinglish, regional tongues and the kind of code-switching that trips many global services. The company says it handled about 30 lakh API calls in the last year - a hint that it’s seen some real-world traffic. In a side-by-side test with Deepgram, Reverie claimed a 1.5× speed edge in voice-agent scenarios, which is impressive if the numbers hold up.
What we don’t get are the nitty-gritty details: how latency behaved on shaky networks, or how transcription error rates varied across accents. The “tuned to India’s real speech” line feels plausible given the call volume, yet without comparative WER figures the actual benefit stays fuzzy. The story also skips over whether the model can scale beyond the pilot cases or plug into existing enterprise pipelines.
So, while the launch marks a genuine attempt to tackle local linguistic quirks, we’ll need clearer benchmarks before saying it truly outperforms the established players.
Further Reading
- Reverie unveils a Voice API Platform Supporting 12 Indian Languages - VAR India
- Reverie Sets A New Benchmark in Automated Speech Recognition (ASR) Accuracy for Indian Native Languages - Business Standard
- Advancing AI Localization: MEDIAWEN and Reverie Expand Video Accessibility in 11 Indian languages - PR Newswire
More in Business & Startups
AI shifts hiring: companies favor generalist engineers over specialists
Why are hiring managers rethinking their talent playbooks? The rise of AI tools is nudging firms...
Rivian showcases new AI chip and silicon inspection labs in office tour
Rivian is opening its doors to a behind‑the‑scenes look at the hardware that could define the next...
Anthropic's AI store profits as CEO acknowledges eternal transcendence
Anthropic’s new AI storefront is finally turning a profit, a milestone that feels almost ceremonial...
Gnani.ai launches Vachana STT model as core IndiaAI infrastructure
Gnani.ai, the Bengaluru‑based speech‑technology firm, has just put its newest speech‑to‑text engine...
PC Company Invests in AI Engines, Context‑Aware Computing, Intelligence
A PC maker long known for assembling laptops and desktops is now positioning itself as a backbone...
Common Questions Answered
What speed advantage does Reverie's new Speech-to-Text (STT) model claim over Deepgram?
Reverie reports that its latest STT model processes audio roughly 1.5 times faster than Deepgram in independent voice‑agent tests. This acceleration can lead to more responsive voice assistants and lower latency for real‑time applications.
How does the Reverie STT model handle India's multilingual reality, including Hinglish and regional dialects?
The model is specifically tuned to Indian speech patterns, capturing code‑switching between Hindi and English as well as the nuances of various regional dialects. By training on authentic Indian audio, it can recognize mixed‑language utterances that often trip up generic global systems.
How many API calls has Reverie's STT engine processed over the past year, and what does this indicate?
Reverie logged 30 lakh (3 million) API calls in the last twelve months, demonstrating substantial real‑world exposure. This volume suggests the model has been tested across diverse use cases and is already being adopted by developers.
Why is an India‑focused speech‑to‑text engine crucial for voice assistants in the Indian market?
India’s linguistic landscape includes dozens of scripts, frequent Hindi‑English code‑switching, and many regional dialects that generic models struggle to decode. A tailored STT engine improves accuracy and user experience for voice assistants operating in this complex environment.