Whissle AI Blog
Explore the latest in AI, machine learning, and voice technology.
Whissle Is a Whistle: Raw Streams In, Live Structured Intell…
A whistle turns a stream of breath into structured sound — instantly, while you're still exhaling. That's exactly what W…
By Karan Singla
Jul 03 2026
Catching Lies Without Sending the Video: Privacy-Preserving …
A widely-cited 75% on this courtroom dataset is inflated by speaker leakage — honest, speaker-independent evaluation is …
By Nikita Sharma, Pranav Saran, Karan Singla
Jun 20 2026
Gujlish META-ASR: A Bilingual English-Gujarati Speech Model …
We open-source a bilingual English-Gujarati speech recognition model that transcribes speech while extracting speaker ag…
By Whissle Research Team
May 16 2026
Building a Hindi Speech Model That Understands Context
We benchmark our META-ASR Hindi model against Deepgram Nova-2 and Gemini 2.5 Flash across two test sets, measuring not j…
By Whissle Research Team
May 03 2026
Introducing Whissle Browser — The First Browser That Reads t…
Today we're launching Whissle Browser, a next-generation browser with Lulu — an ambient AI companion that listens, under…
By Karan Jakhar
May 01 2026
Mandarin ASR Beyond Words: Transcription, Demographics, and …
We benchmarked Whissle's 157M-parameter Chinese model against Deepgram Nova-3 and Gemini 2.5 Flash on 5,200 samples acro…
By Whissle Research Team
Apr 27 2026
Does Your ASR's Metadata Actually Make AI Responses Better? …
613 conversations across three public datasets. Six evaluation dimensions. The metadata-aware pipeline won 86% on conver…
By Whissle Research Team
Apr 20 2026
Beyond Transcription: How a Meta-Aware ASR Model Delivers Wo…
A single CTC model that outputs transcription and metadata action tokens — emotion, intent, speech rate, demographics — …
By Whissle Research Team
Apr 16 2026
We Benchmarked 3 Streaming ASR Providers Across 17 Hours of …
4,915 samples. Four datasets. Clean speech, Indian-accented tech interviews, noisy soccer broadcasts, and Indian Supreme…
By Whissle Research Team
Apr 10 2026
Using Visuals for Better Sound Awareness
Humans naturally use visual cues to understand speech in noisy places. This article explores how we're teaching AI to do…
By Whissle Research Team
Apr 18 2025
Meta-aware Voice Action Model
To create AI companions that feel genuinely interactive, speech recognition must go beyond raw transcription — it must u…
By Whissle Research Team
Mar 16 2025