search

zhouwg / kantv

workbench for learing&practising AI tech in real scenario on Android device, powered by GGML(Georgi Gerganov Machine Learning) and NCNN(Tencent NCNN) and FFmpeg
Apache License 2.0
117 stars 19 forks source link

real real-time transcription(real-time subtitle) with English online-TV on Xiaomi 14 at the first time but buf-fix is still required #86

Closed zhouwg closed 5 months ago

zhouwg commented 5 months ago

another fine-tune which introduced in https://github.com/ggerganov/whisper.cpp/issues/1951.

performance of real-time transcription on Xiaomi14 was improved very significantly

before fine-tune:

Screenshot from 2024-03-16 21-18-24

after fine-tune:

Screenshot from 2024-03-20 16-40-19

but this fine-tune also brings an unexpected side-effect:whispercpp would produce incorrect/repeat tokens or app would crash suddenly.