Incorrect results from transcribing audio from file #740
Description
Description
When developing timestamping in STT I spotted that the results of the STT transcription are a bit off. The first 35 seconds of 58-second audio are recognised as music and audience chattering. Then, transcription is perfectly fine. The audio's quality is the same during the whole listening. The result is presented on the following screenshot:
Reference audio: https://ai.swmansion.com/storage/moonshine/test_audio.mp3
Steps to reproduce
Run STT example from demo app and pass link from the description.
Snack or a link to a repository
No response
React Native Executorch version
Main
React Native version
0.81.5
Platforms
iOS
JavaScript runtime
Hermes
Workflow
None
Architecture
Fabric (New Architecture)
Build type
Debug mode
Device
iOS simulator
Device model
iPhone 16 Pro
AI model
model from STT demo app
Performance logs
No response
Acknowledgements
Yes