Monsoon speech dataset cuts Telugu WER from 92.7% to 16.1%
The takeaway is that low-resource ASR quality can be bought with better data curation rather than bigger models, since a mid-size Whisper checkpoint beat Gemini 3.1 Pro on Telugu purely through fine-tuning. Teams building speech products for Indic and other under-served language markets can use this dataset-first approach instead of waiting for larger architectures. Keeping noisy audio in training, with DNSMOS-based stratification, is the transferable technique.
VoiceArena launched Monsoon, a 50-language speech dataset built from unscripted conversations recorded across 23 countries.
Whisper Medium, off the shelf, scores 92.7% semantic WER on IndicVoices Telugu.
Fine-tuning the same 769M checkpoint on Monsoon drops that to 16.1%, narrowly ahead of MAI-Transcribe-2 and Gemini 3.
1 Pro in VoiceArena's own tests.The 76.
6-point improvement comes entirely from training data, with no architecture change or larger model.Monsoon deliberately keeps noise, short clips and degraded audio in the training mix, stratified by DNSMOS so the share of difficult audio is controlled.
Low-resource ASR (automatic speech recognition) usually gets treated as a model problem.
VoiceArena just launched Monsoon, and shows it's a data collection problem.
Monsoon is a 50-language speech dataset built from unscripted conversations across 23 countries.
Off the shelf, Whisper Medium scores 92.7% semantic WER on IndicVoices Telugu. VoiceArena fine-tuned the same 769M checkpoint on Monsoon and got 16.1%, narrowly ahead of MAI-Transcribe-2 and Gemini 3.1 Pro in its own tests.
There was no architecture change and no bigger model. The 76.6-point drop comes entirely from what the model heard during fine-tuning.
Most speech data pipelines filter noise out. Monsoon keeps it in on purpose.
Short clips, noisy rooms and difficult acoustic conditions stay in the training distribution, and the mix isn't left to chance. VoiceArena stratifies segments by DNSMOS. That makes the share of degraded audio a controlled property of the data
This report is categorized under AITOP's "Models" desk. Within our cross-regional analytical framework, it directly maps to AI developer toolchains, coding agents, and engineering workflows. Key entities and technical topics under continuous tracking include Monsoon, Whisper, VoiceArena, ASR.
Source attribution: curated from rohanpaul_ai (@rohanpaul_ai).
