Sync Space to FunASR 1.4.1

Pin the current public patch release and refresh the release summary.

README.md

@@ -22,7 +22,7 @@ Upload a Chinese-language video → FunASR Paraformer transcribes it with senten
 - Lets you select and concatenate one or more sentences
 - Exports the result as an MP4 with FFmpeg
 
-Short videos provide the best interactive experience on the shared CPU runtime. For multilingual ASR and LLM-assisted smart clipping, install the full [FunClip](https://github.com/modelscope/FunClip) app locally. This Space pins the current public patch, `funasr==1.3.29`; it includes the 1.3.28 real-time final-text fixes and restores SenseVoice VAD regions in `sentence_info` when token timestamps are unavailable. [Release notes](https://github.com/modelscope/FunASR/releases/tag/v1.3.29)
+Short videos provide the best interactive experience on the shared CPU runtime. For multilingual ASR and LLM-assisted smart clipping, install the full [FunClip](https://github.com/modelscope/FunClip) app locally. This Space pins the current public patch, `funasr==1.4.1`; it includes the corrected Hugging Face `paraformer-en` route, safer checkpoint selection, native-runtime ownership and timestamp fixes, and SenseVoice TensorRT/Triton deployment support. [Release notes](https://github.com/modelscope/FunASR/releases/tag/v1.4.1)
 
 ## Links
 - **GitHub**: [Star the FunClip repository](https://github.com/modelscope/FunClip)

requirements.txt

@@ -2,7 +2,7 @@
 
 torch==2.11.0+cpu
 torchaudio==2.11.0+cpu
-funasr==1.3.29
+funasr==1.4.1
 huggingface_hub>=0.33.5,<2.0
 gradio==5.50.0
 numpy<2.0