Realtime

[thewh1teagle]

How should I show the transcibed words in realtime? I want to add this feature to vibe Basiclly I need to register callback function for every chunk which transcribed and display it So I'm wondering, if I use small batch size so user see the progress in small durations, will it affect the performance? Anyway, I appreciate if you can tell me how to implement it or add some example for that, thanks!

[bruceunx]

I implement the following callback to stream realtime data. I checked your repo vibe, It has some version conflict with whisper-rs

static mut COUNT: i32 = 0;

unsafe extern "C" fn callback(
    _ctx: *mut whisper_rs_sys::whisper_context,
    state: *mut whisper_rs_sys::whisper_state,
    _progress: std::os::raw::c_int,
    _user_data: *mut std::ffi::c_void,
) {
    let num_segments = whisper_rs_sys::whisper_full_n_segments_from_state(state);
    for i in COUNT..num_segments {
        let ret = whisper_rs_sys::whisper_full_get_segment_text_from_state(state, i);
        let c_str = CStr::from_ptr(ret);
        let r_str = c_str.to_str().unwrap();
        let t0 = whisper_rs_sys::whisper_full_get_segment_t0_from_state(state, i);
        let t1 = whisper_rs_sys::whisper_full_get_segment_t1_from_state(state, i);
        println!("[{} -> {}]: {}", t0, t1, r_str);
    }

    COUNT = num_segments;
}

    unsafe {
        params.set_progress_callback(Some(callback));
    }