Spaces:
Build error
Build error
| import sys | |
| import time | |
| from pathlib import Path | |
| import anvil.server | |
| import anvil.media | |
| from whisper.utils import write_srt | |
| from youtube_dl import YoutubeDL | |
| from youtube_dl.utils import DownloadError | |
| import os | |
| import tempfile | |
| import json | |
| import whisper | |
| from whisper.tokenizer import LANGUAGES | |
| import ffmpeg | |
| from utils.subs import bake_subs | |
| original_dir = os.getcwd() | |
| output_dir = Path('output') | |
| def download_generator(url): | |
| ### Step 1 : check if video is available | |
| yield {"message": f"Checking {url} for videos"} | |
| try: | |
| meta = check_download(url) | |
| if(meta['duration'] > 5 * 60): | |
| raise Exception("Video is too long, please use videos less than 5 minutes") | |
| yield {"message": f"Found video with {meta['duration']} seconds duration from {meta['extractor']}", "meta": meta} | |
| tempdir = output_dir/f"{meta['id']}" | |
| except Exception as e: | |
| yield {"message": f"{e}"} | |
| return | |
| ### Step 2 : Download video and extract audio | |
| yield {"message": f"Starting download with URL {url}, this may take a while"} | |
| try: | |
| meta, video, audio = download(url, tempdir) | |
| yield {"message": f"Downloaded video and extracted audio", "video": video, "audio": audio, "meta": meta} | |
| except Exception as e: | |
| os.chdir(original_dir) | |
| yield {"message": f"{e}"} | |
| raise e | |
| ### Step 3 : Transcribe with whisper | |
| yield {"message": f"[PLEASE WAIT] Starting whisper transcribe with {meta['id']}.mp3"} | |
| try: | |
| whisper_result = transcribe(audio) | |
| srt_path = tempdir / f"{meta['id']}.srt" | |
| with open(srt_path, "w", encoding="utf-8") as srt: | |
| write_srt(whisper_result["segments"], file=srt) | |
| whisper_result["srt"] = Path(srt_path).read_text() | |
| yield {"message": f"Transcribe successful", "whisper_result": whisper_result, "meta": meta} | |
| except Exception as e: | |
| os.chdir(original_dir) | |
| yield {"message": f"{e}"} | |
| #TODO: add return here so users can continue after editing subtitles | |
| ### Step 4 : Bake subtitles into video with ffmpeg | |
| yield {"message": f"[PLEASE WAIT] baking subtitles into video"} | |
| try: | |
| subbed_video_path = tempdir / f"{meta['id']}_translated.mp4" | |
| fontsdir = Path('fonts') | |
| bake_subs(video, subbed_video_path.absolute() , srt_path.absolute(), fontsdir) | |
| yield {"message": f"Subtitled video ready!", "sub_video": str(subbed_video_path.absolute()), "meta": meta} | |
| except ffmpeg.Error as e: | |
| print('stdout:', e.stdout.decode('utf8')) | |
| print('stderr:', e.stderr.decode('utf8')) | |
| raise e | |
| except Exception as e: | |
| os.chdir(original_dir) | |
| print('error', file=sys.stderr) | |
| raise e | |
| yield {"message": f"{e}"} | |
| def progress_hook(d): | |
| if d['status'] == 'downloading': | |
| print("downloading " + str(round(float(d['downloaded_bytes']) / float(d['total_bytes']) * 100, 1)) + "%") | |
| yield f"{d['_percent_str']} downloaded" | |
| if d['status'] == 'finished': | |
| filename = d['filename'] | |
| print(filename) | |
| yield f"Downloaded {filename}" | |
| def download(url, tempdir): | |
| try: | |
| ydl_opts = { | |
| "format": "bestvideo[ext=mp4]+bestaudio/best", | |
| "keepvideo": True, | |
| 'postprocessors': [{ | |
| 'key': 'FFmpegExtractAudio', | |
| 'preferredcodec': 'mp3', | |
| 'preferredquality': '192', | |
| }], | |
| "skip_download": False, | |
| "outtmpl": f"{tempdir}/%(id)s.%(ext)s", | |
| "noplaylist": True, | |
| "verbose": False, | |
| "quiet": True, | |
| "progress_hooks": [progress_hook], | |
| } | |
| ydl = YoutubeDL(ydl_opts) | |
| meta = ydl.extract_info( | |
| url, | |
| download=True, | |
| ) | |
| except DownloadError as e: | |
| raise e | |
| else: | |
| video = tempdir / f"{meta['id']}.{meta['ext']}" | |
| audio = tempdir / f"{meta['id']}.mp3" | |
| print(str(video.resolve())) | |
| return meta, str(video.resolve()), str(audio.resolve()) | |
| def check_download(url): | |
| ydl_opts = { | |
| "format": "bestvideo[ext=mp4]+bestaudio/best", | |
| "skip_download": True, | |
| "verbose": False, | |
| } | |
| ydl = YoutubeDL(ydl_opts) | |
| try: | |
| meta = ydl.extract_info( | |
| url, | |
| download=False, | |
| ) | |
| except DownloadError as e: | |
| raise e | |
| else: | |
| return meta | |
| def transcribe(audio): | |
| print('Starting transcribe...') | |
| model = whisper.load_model('medium') | |
| output = model.transcribe(audio, task="translate") | |
| output["language"] = LANGUAGES[output["language"]] | |
| # output = {"text": " And then, you throw it and bam! How do you do this when it's going this way? Throw it and bam! That's right! Is that it? Throw it and bam! That's it. That's it. That's it. That's it. That's it. That's right. Throw it and bam! No, this is... Do it like this and... Isn't this the most natural? Bam! What is it?", "segments": [{"id": 0, "seek": 0, "start": 0.0, "end": 2.46, "text": " And then, you throw it and bam!", "tokens": [50364, 400, 550, 11, 291, 3507, 309, 293, 18132, 0, 50487, 50487, 1012, 360, 291, 360, 341, 562, 309, 311, 516, 341, 636, 30, 50607, 50607, 22228, 309, 293, 18132, 0, 50682, 50682, 663, 311, 558, 0, 50760, 50760, 1119, 300, 309, 30, 50785, 50785, 22228, 309, 293, 18132, 0, 50852, 50852, 663, 311, 309, 13, 50877, 50877, 663, 311, 309, 13, 50902, 50902, 663, 311, 309, 13, 50927, 50927, 663, 311, 309, 13, 50952, 50952, 663, 311, 309, 13, 50977, 50977, 663, 311, 558, 13, 51002, 51002, 22228, 309, 293, 18132, 0, 51094, 51094, 883, 11, 341, 307, 485, 51130, 51130, 1144, 309, 411, 341, 293, 485, 51182, 51182, 6998, 380, 341, 264, 881, 3303, 30, 51260, 51260, 26630, 0, 51310, 51310, 663, 311, 309, 13, 51354], "temperature": 0.0, "avg_logprob": -0.8081162918445676, "compression_ratio": 2.0533333333333332, "no_speech_prob": 0.14013107120990753}, {"id": 1, "seek": 0, "start": 2.46, "end": 4.86, "text": " How do you do this when it's going this way?", "tokens": [50364, 400, 550, 11, 291, 3507, 309, 293, 18132, 0, 50487, 50487, 1012, 360, 291, 360, 341, 562, 309, 311, 516, 341, 636, 30, 50607, 50607, 22228, 309, 293, 18132, 0, 50682, 50682, 663, 311, 558, 0, 50760, 50760, 1119, 300, 309, 30, 50785, 50785, 22228, 309, 293, 18132, 0, 50852, 50852, 663, 311, 309, 13, 50877, 50877, 663, 311, 309, 13, 50902, 50902, 663, 311, 309, 13, 50927, 50927, 663, 311, 309, 13, 50952, 50952, 663, 311, 309, 13, 50977, 50977, 663, 311, 558, 13, 51002, 51002, 22228, 309, 293, 18132, 0, 51094, 51094, 883, 11, 341, 307, 485, 51130, 51130, 1144, 309, 411, 341, 293, 485, 51182, 51182, 6998, 380, 341, 264, 881, 3303, 30, 51260, 51260, 26630, 0, 51310, 51310, 663, 311, 309, 13, 51354], "temperature": 0.0, "avg_logprob": -0.8081162918445676, "compression_ratio": 2.0533333333333332, "no_speech_prob": 0.14013107120990753}, {"id": 2, "seek": 0, "start": 4.86, "end": 6.36, "text": " Throw it and bam!", "tokens": [50364, 400, 550, 11, 291, 3507, 309, 293, 18132, 0, 50487, 50487, 1012, 360, 291, 360, 341, 562, 309, 311, 516, 341, 636, 30, 50607, 50607, 22228, 309, 293, 18132, 0, 50682, 50682, 663, 311, 558, 0, 50760, 50760, 1119, 300, 309, 30, 50785, 50785, 22228, 309, 293, 18132, 0, 50852, 50852, 663, 311, 309, 13, 50877, 50877, 663, 311, 309, 13, 50902, 50902, 663, 311, 309, 13, 50927, 50927, 663, 311, 309, 13, 50952, 50952, 663, 311, 309, 13, 50977, 50977, 663, 311, 558, 13, 51002, 51002, 22228, 309, 293, 18132, 0, 51094, 51094, 883, 11, 341, 307, 485, 51130, 51130, 1144, 309, 411, 341, 293, 485, 51182, 51182, 6998, 380, 341, 264, 881, 3303, 30, 51260, 51260, 26630, 0, 51310, 51310, 663, 311, 309, 13, 51354], "temperature": 0.0, "avg_logprob": -0.8081162918445676, "compression_ratio": 2.0533333333333332, "no_speech_prob": 0.14013107120990753}, {"id": 3, "seek": 0, "start": 6.36, "end": 7.92, "text": " That's right!", "tokens": [50364, 400, 550, 11, 291, 3507, 309, 293, 18132, 0, 50487, 50487, 1012, 360, 291, 360, 341, 562, 309, 311, 516, 341, 636, 30, 50607, 50607, 22228, 309, 293, 18132, 0, 50682, 50682, 663, 311, 558, 0, 50760, 50760, 1119, 300, 309, 30, 50785, 50785, 22228, 309, 293, 18132, 0, 50852, 50852, 663, 311, 309, 13, 50877, 50877, 663, 311, 309, 13, 50902, 50902, 663, 311, 309, 13, 50927, 50927, 663, 311, 309, 13, 50952, 50952, 663, 311, 309, 13, 50977, 50977, 663, 311, 558, 13, 51002, 51002, 22228, 309, 293, 18132, 0, 51094, 51094, 883, 11, 341, 307, 485, 51130, 51130, 1144, 309, 411, 341, 293, 485, 51182, 51182, 6998, 380, 341, 264, 881, 3303, 30, 51260, 51260, 26630, 0, 51310, 51310, 663, 311, 309, 13, 51354], "temperature": 0.0, "avg_logprob": -0.8081162918445676, "compression_ratio": 2.0533333333333332, "no_speech_prob": 0.14013107120990753}, {"id": 4, "seek": 0, "start": 7.92, "end": 8.42, "text": " Is that it?", "tokens": [50364, 400, 550, 11, 291, 3507, 309, 293, 18132, 0, 50487, 50487, 1012, 360, 291, 360, 341, 562, 309, 311, 516, 341, 636, 30, 50607, 50607, 22228, 309, 293, 18132, 0, 50682, 50682, 663, 311, 558, 0, 50760, 50760, 1119, 300, 309, 30, 50785, 50785, 22228, 309, 293, 18132, 0, 50852, 50852, 663, 311, 309, 13, 50877, 50877, 663, 311, 309, 13, 50902, 50902, 663, 311, 309, 13, 50927, 50927, 663, 311, 309, 13, 50952, 50952, 663, 311, 309, 13, 50977, 50977, 663, 311, 558, 13, 51002, 51002, 22228, 309, 293, 18132, 0, 51094, 51094, 883, 11, 341, 307, 485, 51130, 51130, 1144, 309, 411, 341, 293, 485, 51182, 51182, 6998, 380, 341, 264, 881, 3303, 30, 51260, 51260, 26630, 0, 51310, 51310, 663, 311, 309, 13, 51354], "temperature": 0.0, "avg_logprob": -0.8081162918445676, "compression_ratio": 2.0533333333333332, "no_speech_prob": 0.14013107120990753}, {"id": 5, "seek": 0, "start": 8.42, "end": 9.76, "text": " Throw it and bam!", "tokens": [50364, 400, 550, 11, 291, 3507, 309, 293, 18132, 0, 50487, 50487, 1012, 360, 291, 360, 341, 562, 309, 311, 516, 341, 636, 30, 50607, 50607, 22228, 309, 293, 18132, 0, 50682, 50682, 663, 311, 558, 0, 50760, 50760, 1119, 300, 309, 30, 50785, 50785, 22228, 309, 293, 18132, 0, 50852, 50852, 663, 311, 309, 13, 50877, 50877, 663, 311, 309, 13, 50902, 50902, 663, 311, 309, 13, 50927, 50927, 663, 311, 309, 13, 50952, 50952, 663, 311, 309, 13, 50977, 50977, 663, 311, 558, 13, 51002, 51002, 22228, 309, 293, 18132, 0, 51094, 51094, 883, 11, 341, 307, 485, 51130, 51130, 1144, 309, 411, 341, 293, 485, 51182, 51182, 6998, 380, 341, 264, 881, 3303, 30, 51260, 51260, 26630, 0, 51310, 51310, 663, 311, 309, 13, 51354], "temperature": 0.0, "avg_logprob": -0.8081162918445676, "compression_ratio": 2.0533333333333332, "no_speech_prob": 0.14013107120990753}, {"id": 6, "seek": 0, "start": 9.76, "end": 10.26, "text": " That's it.", "tokens": [50364, 400, 550, 11, 291, 3507, 309, 293, 18132, 0, 50487, 50487, 1012, 360, 291, 360, 341, 562, 309, 311, 516, 341, 636, 30, 50607, 50607, 22228, 309, 293, 18132, 0, 50682, 50682, 663, 311, 558, 0, 50760, 50760, 1119, 300, 309, 30, 50785, 50785, 22228, 309, 293, 18132, 0, 50852, 50852, 663, 311, 309, 13, 50877, 50877, 663, 311, 309, 13, 50902, 50902, 663, 311, 309, 13, 50927, 50927, 663, 311, 309, 13, 50952, 50952, 663, 311, 309, 13, 50977, 50977, 663, 311, 558, 13, 51002, 51002, 22228, 309, 293, 18132, 0, 51094, 51094, 883, 11, 341, 307, 485, 51130, 51130, 1144, 309, 411, 341, 293, 485, 51182, 51182, 6998, 380, 341, 264, 881, 3303, 30, 51260, 51260, 26630, 0, 51310, 51310, 663, 311, 309, 13, 51354], "temperature": 0.0, "avg_logprob": -0.8081162918445676, "compression_ratio": 2.0533333333333332, "no_speech_prob": 0.14013107120990753}, {"id": 7, "seek": 0, "start": 10.26, "end": 10.76, "text": " That's it.", "tokens": [50364, 400, 550, 11, 291, 3507, 309, 293, 18132, 0, 50487, 50487, 1012, 360, 291, 360, 341, 562, 309, 311, 516, 341, 636, 30, 50607, 50607, 22228, 309, 293, 18132, 0, 50682, 50682, 663, 311, 558, 0, 50760, 50760, 1119, 300, 309, 30, 50785, 50785, 22228, 309, 293, 18132, 0, 50852, 50852, 663, 311, 309, 13, 50877, 50877, 663, 311, 309, 13, 50902, 50902, 663, 311, 309, 13, 50927, 50927, 663, 311, 309, 13, 50952, 50952, 663, 311, 309, 13, 50977, 50977, 663, 311, 558, 13, 51002, 51002, 22228, 309, 293, 18132, 0, 51094, 51094, 883, 11, 341, 307, 485, 51130, 51130, 1144, 309, 411, 341, 293, 485, 51182, 51182, 6998, 380, 341, 264, 881, 3303, 30, 51260, 51260, 26630, 0, 51310, 51310, 663, 311, 309, 13, 51354], "temperature": 0.0, "avg_logprob": -0.8081162918445676, "compression_ratio": 2.0533333333333332, "no_speech_prob": 0.14013107120990753}, {"id": 8, "seek": 0, "start": 10.76, "end": 11.26, "text": " That's it.", "tokens": [50364, 400, 550, 11, 291, 3507, 309, 293, 18132, 0, 50487, 50487, 1012, 360, 291, 360, 341, 562, 309, 311, 516, 341, 636, 30, 50607, 50607, 22228, 309, 293, 18132, 0, 50682, 50682, 663, 311, 558, 0, 50760, 50760, 1119, 300, 309, 30, 50785, 50785, 22228, 309, 293, 18132, 0, 50852, 50852, 663, 311, 309, 13, 50877, 50877, 663, 311, 309, 13, 50902, 50902, 663, 311, 309, 13, 50927, 50927, 663, 311, 309, 13, 50952, 50952, 663, 311, 309, 13, 50977, 50977, 663, 311, 558, 13, 51002, 51002, 22228, 309, 293, 18132, 0, 51094, 51094, 883, 11, 341, 307, 485, 51130, 51130, 1144, 309, 411, 341, 293, 485, 51182, 51182, 6998, 380, 341, 264, 881, 3303, 30, 51260, 51260, 26630, 0, 51310, 51310, 663, 311, 309, 13, 51354], "temperature": 0.0, "avg_logprob": -0.8081162918445676, "compression_ratio": 2.0533333333333332, "no_speech_prob": 0.14013107120990753}, {"id": 9, "seek": 0, "start": 11.26, "end": 11.76, "text": " That's it.", "tokens": [50364, 400, 550, 11, 291, 3507, 309, 293, 18132, 0, 50487, 50487, 1012, 360, 291, 360, 341, 562, 309, 311, 516, 341, 636, 30, 50607, 50607, 22228, 309, 293, 18132, 0, 50682, 50682, 663, 311, 558, 0, 50760, 50760, 1119, 300, 309, 30, 50785, 50785, 22228, 309, 293, 18132, 0, 50852, 50852, 663, 311, 309, 13, 50877, 50877, 663, 311, 309, 13, 50902, 50902, 663, 311, 309, 13, 50927, 50927, 663, 311, 309, 13, 50952, 50952, 663, 311, 309, 13, 50977, 50977, 663, 311, 558, 13, 51002, 51002, 22228, 309, 293, 18132, 0, 51094, 51094, 883, 11, 341, 307, 485, 51130, 51130, 1144, 309, 411, 341, 293, 485, 51182, 51182, 6998, 380, 341, 264, 881, 3303, 30, 51260, 51260, 26630, 0, 51310, 51310, 663, 311, 309, 13, 51354], "temperature": 0.0, "avg_logprob": -0.8081162918445676, "compression_ratio": 2.0533333333333332, "no_speech_prob": 0.14013107120990753}, {"id": 10, "seek": 0, "start": 11.76, "end": 12.26, "text": " That's it.", "tokens": [50364, 400, 550, 11, 291, 3507, 309, 293, 18132, 0, 50487, 50487, 1012, 360, 291, 360, 341, 562, 309, 311, 516, 341, 636, 30, 50607, 50607, 22228, 309, 293, 18132, 0, 50682, 50682, 663, 311, 558, 0, 50760, 50760, 1119, 300, 309, 30, 50785, 50785, 22228, 309, 293, 18132, 0, 50852, 50852, 663, 311, 309, 13, 50877, 50877, 663, 311, 309, 13, 50902, 50902, 663, 311, 309, 13, 50927, 50927, 663, 311, 309, 13, 50952, 50952, 663, 311, 309, 13, 50977, 50977, 663, 311, 558, 13, 51002, 51002, 22228, 309, 293, 18132, 0, 51094, 51094, 883, 11, 341, 307, 485, 51130, 51130, 1144, 309, 411, 341, 293, 485, 51182, 51182, 6998, 380, 341, 264, 881, 3303, 30, 51260, 51260, 26630, 0, 51310, 51310, 663, 311, 309, 13, 51354], "temperature": 0.0, "avg_logprob": -0.8081162918445676, "compression_ratio": 2.0533333333333332, "no_speech_prob": 0.14013107120990753}, {"id": 11, "seek": 0, "start": 12.26, "end": 12.76, "text": " That's right.", "tokens": [50364, 400, 550, 11, 291, 3507, 309, 293, 18132, 0, 50487, 50487, 1012, 360, 291, 360, 341, 562, 309, 311, 516, 341, 636, 30, 50607, 50607, 22228, 309, 293, 18132, 0, 50682, 50682, 663, 311, 558, 0, 50760, 50760, 1119, 300, 309, 30, 50785, 50785, 22228, 309, 293, 18132, 0, 50852, 50852, 663, 311, 309, 13, 50877, 50877, 663, 311, 309, 13, 50902, 50902, 663, 311, 309, 13, 50927, 50927, 663, 311, 309, 13, 50952, 50952, 663, 311, 309, 13, 50977, 50977, 663, 311, 558, 13, 51002, 51002, 22228, 309, 293, 18132, 0, 51094, 51094, 883, 11, 341, 307, 485, 51130, 51130, 1144, 309, 411, 341, 293, 485, 51182, 51182, 6998, 380, 341, 264, 881, 3303, 30, 51260, 51260, 26630, 0, 51310, 51310, 663, 311, 309, 13, 51354], "temperature": 0.0, "avg_logprob": -0.8081162918445676, "compression_ratio": 2.0533333333333332, "no_speech_prob": 0.14013107120990753}, {"id": 12, "seek": 0, "start": 12.76, "end": 14.6, "text": " Throw it and bam!", "tokens": [50364, 400, 550, 11, 291, 3507, 309, 293, 18132, 0, 50487, 50487, 1012, 360, 291, 360, 341, 562, 309, 311, 516, 341, 636, 30, 50607, 50607, 22228, 309, 293, 18132, 0, 50682, 50682, 663, 311, 558, 0, 50760, 50760, 1119, 300, 309, 30, 50785, 50785, 22228, 309, 293, 18132, 0, 50852, 50852, 663, 311, 309, 13, 50877, 50877, 663, 311, 309, 13, 50902, 50902, 663, 311, 309, 13, 50927, 50927, 663, 311, 309, 13, 50952, 50952, 663, 311, 309, 13, 50977, 50977, 663, 311, 558, 13, 51002, 51002, 22228, 309, 293, 18132, 0, 51094, 51094, 883, 11, 341, 307, 485, 51130, 51130, 1144, 309, 411, 341, 293, 485, 51182, 51182, 6998, 380, 341, 264, 881, 3303, 30, 51260, 51260, 26630, 0, 51310, 51310, 663, 311, 309, 13, 51354], "temperature": 0.0, "avg_logprob": -0.8081162918445676, "compression_ratio": 2.0533333333333332, "no_speech_prob": 0.14013107120990753}, {"id": 13, "seek": 0, "start": 14.6, "end": 15.32, "text": " No, this is...", "tokens": [50364, 400, 550, 11, 291, 3507, 309, 293, 18132, 0, 50487, 50487, 1012, 360, 291, 360, 341, 562, 309, 311, 516, 341, 636, 30, 50607, 50607, 22228, 309, 293, 18132, 0, 50682, 50682, 663, 311, 558, 0, 50760, 50760, 1119, 300, 309, 30, 50785, 50785, 22228, 309, 293, 18132, 0, 50852, 50852, 663, 311, 309, 13, 50877, 50877, 663, 311, 309, 13, 50902, 50902, 663, 311, 309, 13, 50927, 50927, 663, 311, 309, 13, 50952, 50952, 663, 311, 309, 13, 50977, 50977, 663, 311, 558, 13, 51002, 51002, 22228, 309, 293, 18132, 0, 51094, 51094, 883, 11, 341, 307, 485, 51130, 51130, 1144, 309, 411, 341, 293, 485, 51182, 51182, 6998, 380, 341, 264, 881, 3303, 30, 51260, 51260, 26630, 0, 51310, 51310, 663, 311, 309, 13, 51354], "temperature": 0.0, "avg_logprob": -0.8081162918445676, "compression_ratio": 2.0533333333333332, "no_speech_prob": 0.14013107120990753}, {"id": 14, "seek": 0, "start": 15.32, "end": 16.36, "text": " Do it like this and...", "tokens": [50364, 400, 550, 11, 291, 3507, 309, 293, 18132, 0, 50487, 50487, 1012, 360, 291, 360, 341, 562, 309, 311, 516, 341, 636, 30, 50607, 50607, 22228, 309, 293, 18132, 0, 50682, 50682, 663, 311, 558, 0, 50760, 50760, 1119, 300, 309, 30, 50785, 50785, 22228, 309, 293, 18132, 0, 50852, 50852, 663, 311, 309, 13, 50877, 50877, 663, 311, 309, 13, 50902, 50902, 663, 311, 309, 13, 50927, 50927, 663, 311, 309, 13, 50952, 50952, 663, 311, 309, 13, 50977, 50977, 663, 311, 558, 13, 51002, 51002, 22228, 309, 293, 18132, 0, 51094, 51094, 883, 11, 341, 307, 485, 51130, 51130, 1144, 309, 411, 341, 293, 485, 51182, 51182, 6998, 380, 341, 264, 881, 3303, 30, 51260, 51260, 26630, 0, 51310, 51310, 663, 311, 309, 13, 51354], "temperature": 0.0, "avg_logprob": -0.8081162918445676, "compression_ratio": 2.0533333333333332, "no_speech_prob": 0.14013107120990753}, {"id": 15, "seek": 0, "start": 16.36, "end": 17.92, "text": " Isn't this the most natural?", "tokens": [50364, 400, 550, 11, 291, 3507, 309, 293, 18132, 0, 50487, 50487, 1012, 360, 291, 360, 341, 562, 309, 311, 516, 341, 636, 30, 50607, 50607, 22228, 309, 293, 18132, 0, 50682, 50682, 663, 311, 558, 0, 50760, 50760, 1119, 300, 309, 30, 50785, 50785, 22228, 309, 293, 18132, 0, 50852, 50852, 663, 311, 309, 13, 50877, 50877, 663, 311, 309, 13, 50902, 50902, 663, 311, 309, 13, 50927, 50927, 663, 311, 309, 13, 50952, 50952, 663, 311, 309, 13, 50977, 50977, 663, 311, 558, 13, 51002, 51002, 22228, 309, 293, 18132, 0, 51094, 51094, 883, 11, 341, 307, 485, 51130, 51130, 1144, 309, 411, 341, 293, 485, 51182, 51182, 6998, 380, 341, 264, 881, 3303, 30, 51260, 51260, 26630, 0, 51310, 51310, 663, 311, 309, 13, 51354], "temperature": 0.0, "avg_logprob": -0.8081162918445676, "compression_ratio": 2.0533333333333332, "no_speech_prob": 0.14013107120990753}, {"id": 16, "seek": 0, "start": 17.92, "end": 18.92, "text": " Bam!", "tokens": [50364, 400, 550, 11, 291, 3507, 309, 293, 18132, 0, 50487, 50487, 1012, 360, 291, 360, 341, 562, 309, 311, 516, 341, 636, 30, 50607, 50607, 22228, 309, 293, 18132, 0, 50682, 50682, 663, 311, 558, 0, 50760, 50760, 1119, 300, 309, 30, 50785, 50785, 22228, 309, 293, 18132, 0, 50852, 50852, 663, 311, 309, 13, 50877, 50877, 663, 311, 309, 13, 50902, 50902, 663, 311, 309, 13, 50927, 50927, 663, 311, 309, 13, 50952, 50952, 663, 311, 309, 13, 50977, 50977, 663, 311, 558, 13, 51002, 51002, 22228, 309, 293, 18132, 0, 51094, 51094, 883, 11, 341, 307, 485, 51130, 51130, 1144, 309, 411, 341, 293, 485, 51182, 51182, 6998, 380, 341, 264, 881, 3303, 30, 51260, 51260, 26630, 0, 51310, 51310, 663, 311, 309, 13, 51354], "temperature": 0.0, "avg_logprob": -0.8081162918445676, "compression_ratio": 2.0533333333333332, "no_speech_prob": 0.14013107120990753}, {"id": 17, "seek": 1892, "start": 18.92, "end": 30.92, "text": " What is it?", "tokens": [50364, 708, 307, 309, 30, 50964], "temperature": 0.0, "avg_logprob": -0.9666390419006348, "compression_ratio": 0.5789473684210527, "no_speech_prob": 0.0033069916535168886}], "language": "korean"} | |
| output['segments'] = [{"id": 0, "seek": 0, "start": 0.0, "end": 3, "text": " [AI translation by @vidtranslator]"}] + output['segments'] | |
| print(f'Finished transcribe from {output["language"]}', output["text"]) | |
| return output | |