Update README.md

README.md

@@ -76,7 +76,7 @@ pipe = AutomaticSpeechRecognitionPipeline(model=model, feature_extractor=process
 transcript = pipe("audio_samples/short_rd.wav", return_timestamps=False)
 print(transcript)
 ```
-/!\ As of April 10, 2024, the transformers ASR pipeline doesn't allow predicting word timestamps without generating timestamps, resulting in very unreliable timestamps. It is recommended not to attempt predicting timestamps using this pipeline and model.
+
 openai-whisper:
 ```python
 import whisper
@@ -102,10 +102,7 @@ segments, info = model.transcribe("long_audio.wav",
 
 ```
 
-Whisper.cpp: works out-of-the-box
-
+Whisper.cpp:
 ```bash
  ./main -f long_audio.wav -l fr -mc 0 -m ggml-model.bin
-```
- 
-# TODO: Insérer tableau de perfs + liens pour les modèles convertis
+```