arxiv:2505.08699
George Saon
gsaon
AI & ML interests
ASR
Recent Activity
new activity
1 day ago
ibm-granite/granite-speech-3.3-8b:Incomplete Transcripts (with Transformers package)
upvoted
a
collection
13 days ago
Granite Speech
replied to
randomblock1's
post
13 days ago
Try IBM's Granite Speech 3.3 8B in a Space! Currently ranks #2 on the Open ASR Leaderboard (https://huggingface.co/spaces/hf-audio/open_asr_leaderboard) by Word Error Rate.
My go-to transcription model is probably still https://huggingface.co/nvidia/parakeet-tdt-0.6b-v2, as Granite does not perform punctuation or capitalization. Still interesting nonetheless!
https://huggingface.co/spaces/randomblock1/granite-speech-3.3 (sorry it's slow, it runs on base CPU)
Organizations
