merve#674
Second big release from us today: Nemotron-3.5-ASR-Streaming! 🌎40 languages ⚡️80ms - 1s controllable latency 🔥240 - 2400 concurrent streams on 1xH100 🧱FastConformer Cache-Aware RNN-T architecture
https://huggingface.co/nvidia/nemotron-3.5-asr-streaming-0.6b
merve#674Second big release from us today: Nemotron-3.5-ASR-Streaming! 🌎40 languages ⚡️80ms - 1s controllable latency 🔥240 - 2400 concurrent streams on 1xH100 🧱FastConformer Cache-Aware RNN-T architecture
https://huggingface.co/nvidia/nemotron-3.5-asr-streaming-0.6b
Users are praising NVIDIA's Nemotron-3.5-ASR-Streaming release with thanks and excitement because the team appears to be advancing high-quality multilingual ASR models effectively.
@PiotrZelasko Nice! Sadly, my language isn't covered. I just quickly skimmed through the fine-tuning guide,will fine-tuning work for a language that wasn't seen in the training data?
@TrelisResearch For most languages offline Parakeet is going to be more accurate, offline-streaming WER gap is real.
@acul_SR Generally yeah, usually using a pretrained model is better than random initialization for new language expansion, see our newly added fine-tuning skill for best practices https://github.com/NVIDIA-NeMo/NeMo/blob/main/.claude/skills/nemo-speech-asr-finetune/SKILL.md
@PiotrZelasko Does it support keyword biasing?