Update README.md

README.md

@@ -14,21 +14,26 @@ license: cc-by-4.0
 
 ## Open Whisper-style Speech Model (OWSM)
 
-OWSM aims to develop fully open speech foundation models using publicly available data and open-source toolkits including [ESPnet](https://github.com/espnet/espnet).
+OWSM aims to develop fully open speech foundation models using publicly available data and open-source toolkits, including [ESPnet](https://github.com/espnet/espnet).
 
-Inference examples can be found in our [project page](https://www.wavlab.org/activities/2024/owsm/).
+Inference examples can be found on our [project page](https://www.wavlab.org/activities/2024/owsm/).
 The Gradio demo is [here](https://huggingface.co/spaces/pyf98/OWSM_v3_demo). 
 
 [OWSM v4]() is the latest version in the OWSM series, which significantly outperforms OWSM v3.1 in LID and multilingual ASR.
+Additionally, OWSM v4 applies 8 times subsampling (instead of 4 times in OWSM v3.1) to the log Mel features, leading to a final resolution of 80 ms in the encoder.
+When running inference, we recommend setting `maxlenratio=1.0` (default) instead of smaller values.
 
-This repo contains a medium-sized model with 1.02B parameters, developed by [Yifan Peng](https://pyf98.github.io/) (CMU). It is trained on 320k hours of public speech data. The newly curated data will be publicly released. Please stay tuned!
+This repo contains a base-sized model with 102M parameters, developed by [Yifan Peng](https://pyf98.github.io/) (CMU). 
+It is trained on 320k hours of public speech data. 
+The newly curated data will be publicly released. Please stay tuned!
 
 It supports the following speech-to-text tasks:
 - Language identification
 - Speech recognition
 - Speech translation
 - Utterance-level timestamp prediction
-- Long-form transcription
+- Long-form recognition or translation
+
 
 
 ### OWSM series