upload.py

import json
from google.cloud import speech
from google.cloud.speech import enums
from google.cloud.speech import types
from pydub import AudioSegment
import pickle

content = None
frames = None
# read frame data

def process_clip():

    client = speech.SpeechClient()

    audio = types.RecognitionAudio(uri='gs://audio-hackprinceton19/video.wav')
    config = types.RecognitionConfig(
        encoding = enums.RecognitionConfig.AudioEncoding.LINEAR16,
        sample_rate_hertz = 48000,
        language_code='en-US',
        enable_word_time_offsets = True)
    response = client.long_running_recognize(config=config, audio=audio).result()
    return response.results

def get_frames(interval):
    trans_results = process_clip()
    content = []
    for i in range(len(trans_results)):
        content.append(trans_results[i].alternatives[0].words)
    content = [(content[i][j].start_time.seconds, content[i][j].start_time.nanos, content[i][j].end_time.seconds,
                content[i][j].end_time.nanos, content[i][j].word) for i in range(len(content)) for j in
               range(len(content[i]))]

    data = content
    emotions = ("joy", "sorrow", "surprise", "anger")
    with open("data.json", "r") as f:
        frames = json.load(f)
    for i, frame in enumerate(frames):
        frames[i]["emotions"] = {}
        for emotion in emotions:
            frames[i]["emotions"][emotion] = frame[emotion]
    frames = [{k: v for k, v in frames[i].items() if k not in emotions} for i in range(len(frames))]
    start, finish = 0, interval
    data_counter = 0
    for i, frame in enumerate(frames):
        text = ""
        if data_counter >= len(data):
            break
        time = data[data_counter][2] + (data[data_counter][3] / (10**9))
        while time <= finish and data_counter < len(data):
            text += (data[data_counter][4] + " ")
            time = data[data_counter][2] + (data[data_counter][3] / (10 ** 9))
            data_counter += 1
        start += interval
        finish += interval
        frames[i]["text"] = text
    return frames


def upload():

    global content, frames

    with open("data.json") as f:
        frames = json.load(f)

    audio_filepath = "./video.wav"
    sound = AudioSegment.from_wav(audio_filepath)
    sound = sound.set_channels(1)
    sound.export(audio_filepath, format="wav")

    trans_results = process_clip(audio_filepath)
    content = []
    for i in range(len(trans_results)):
        content.append(trans_results[i].alternatives[0].words)
    content = [(content[i][j].start_time.seconds, content[i][j].start_time.nanos, content[i][j].end_time.seconds, content[i][j].end_time.nanos, content[i][j].word) for i in range(len(content)) for j in range(len(content[i]))]

    with open("test.pickle", "wb") as f:
        pickle.dump(content, f)
    with open("lmao.txt", "w") as file:
        file.write(str(trans_results))

if __name__  == '__main__':
    #upload()
    get_frames(5)