Initial commit, let's see if this works on HF

README.md

@@ -1,6 +1,6 @@
 ---
 title: Vidtranslator
-emoji: 🐨
+emoji: 🎥
 colorFrom: red
 colorTo: purple
 sdk: gradio

app.py

@@ -1,7 +1,120 @@
 import gradio as gr
 
-def greet(name):
-    return "Hello there " + name + "!!"
+from download import check_download, download, download_generator
 
-iface = gr.Interface(fn=greet, inputs="text", outputs="text")
-iface.launch()
+import anvil.media
+import os
+import pathlib
+from pathlib import Path
+from shutil import rmtree
+
+anvil.server.connect(os.environ['ANVIL_UPLINK_KEY'])
+queue_placeholder = None
+
+url_input = gr.Textbox(label="Youtube/Twitter/etc video URL (supports many services)", value='https://twitter.com/starsonxh/status/1552945347194142720', lines=1, elem_id="url_input")
+download_status = gr.Textbox(label="Status:", value='', lines=1, elem_id="download_status")
+init_video = gr.Video(label="Downloaded video", visible=False)
+init_audio = gr.Audio(label="Downloaded audio", visible=False)
+output_text = gr.Textbox(label="Output text", lines=5, visible=False, max_lines=10)
+sub_video = gr.Video(label="Subbed video", visible=False)
+
+
+@anvil.server.callable
+def cleanup_output_dir():
+  #make sure we're in the main directory
+  os.chdir(pathlib.Path(__file__).parent.absolute())
+  #delete the output directory contents
+  for path in Path("output").glob("**/*"):
+    if path.is_file():
+      path.unlink()
+    elif path.is_dir():
+      rmtree(path)
+
+@anvil.server.callable
+def download_api(url):
+  print(f'Request from Anvil with URL {url}')
+  final_response = ''
+  #TODO: figure out how to push an incoming event to the queue
+  #THIS DOESN'T WORK queue_placeholder.push_event('download', url)
+  #TODO: handle errors
+  for response in download_generator(url):
+    final_response = response['message']
+    print(final_response)
+  return final_response
+
+def predownload(url):
+  for response in download_generator(url):
+    updates_object = {}
+    updates_object[download_status] = response.get('message', '')
+    meta = response.get('meta')
+    if 'video' in response:
+      updates_object[init_video] = gr.update(visible=True, value=response["video"],
+                                             label=f"Init Video: {meta['id']}.{meta['ext']}")
+      updates_object[init_audio] = gr.update(visible=True, value=response["audio"],
+                                             label=f"Extracted audio : {meta['id']}.mp3")
+    if 'whisper_result' in response:
+      updates_object[output_text] = gr.update(value=response['whisper_result'].get('srt'), visible=True,
+                                              label=f"Subtitles translated from {response['whisper_result'].get('language')} (detected language)")
+    if 'sub_video' in response:
+      updates_object[sub_video] = gr.update(visible=True, value=response["sub_video"],
+                                            label=f"Subbed video: {meta['id']}_translated.mp4")
+
+    yield updates_object
+
+
+subtitled_video = False
+css = """
+#submit{
+  position: absolute;
+  flex:0 !important;
+  width: 120px;
+  right: 13px;
+  top: 40px;
+}
+#url_input{
+  font-size: 40px !important;
+}
+#download_status{
+  font-size: 40px !important;
+  }
+.gradio-container {background-color: red}
+#input_row{
+  position: relative;
+}
+.gradio-interface #submit{
+  
+}
+"""
+with gr.Blocks(css=css+"") as demo:
+  gr.Markdown('# Vid Translator 0.1 - get english subtitles for videos in any language')
+  gr.Markdown('### Link to a tweet, youtube or other video and get a translated video with @openAi #whisper, built by [@altryne](https://twitter.com/altryne/)')
+  gr.Markdown('### This is used as the backend for [@vidtranslator](https://twitter.com/vidtranslator/)')
+  with gr.Row(elem_id="input_row"):
+    with gr.Group() as group:
+      url_input.render()
+      greet_btn = gr.Button("Download", elem_id='submit', variant='primary')
+      pause_for_editing = gr.Checkbox(label="Pause for editing")
+  with gr.Row():
+    with gr.Column():
+      download_status.render()
+      init_video.render()
+      init_audio.render()
+    with gr.Column():
+      with gr.Group() :
+        output_text.render()
+        gr.Button("Download srt file")
+        gr.Button("Bake subtitles into video")
+      sub_video.render()
+      if(subtitled_video):
+        download_video = gr.Button("Download Video", variant='primary')
+        download_srt = gr.Button("Download Srt", variant='primary')
+
+  greet_btn.click(fn=predownload, inputs=[url_input], outputs=[download_status, init_video, init_audio, output_text, sub_video], api_name='predownload')
+  url_input.submit(fn=predownload, inputs=[url_input], outputs=[download_status, init_video, init_audio, output_text, sub_video])
+
+  api_button = gr.Button("API", variant='primary', visible=False).click(fn=cleanup_output_dir, inputs=[], outputs=[], api_name='cleanup_output_dir')
+
+queue_placeholder = demo.queue()
+
+if __name__ == "__main__":
+  demo.launch(show_error=True, debug=True)

download.py

@@ -0,0 +1,146 @@
+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)
+    yield {"message": f"Found video with {meta['duration']} seconds duration from {meta['extractor']}", "meta": meta}
+    # create a temp directory with meta ID
+    # os.makedirs(output_dir/f"{meta['id']}", exist_ok=False)
+    tempdir = output_dir/f"{meta['id']}"
+  except Exception as e:
+    os.chdir(original_dir)
+    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"}
+  print(f'Finished transcribe from {output["language"]}', output["text"])
+  return output

utils/subs.py

@@ -0,0 +1,84 @@
+import ffmpeg
+import os
+from pathlib import Path, PureWindowsPath
+import anvil.media
+import os
+from typing import Iterator, TextIO
+
+
+
+def bake_subs(input_file, output_file, subs_file, fontsdir):
+  print(f"Baking {subs_file} into video... {input_file} -> {output_file}")
+
+
+  fontfile = fontsdir / 'arial.ttf'
+  fontstyle = 'Fontsize=18,OutlineColour=&H40000000,BorderStyle=3,FontName=Arial'
+  video = ffmpeg.input(input_file)
+  audio = video.audio
+  (
+    ffmpeg
+      .concat(
+      video.filter('subtitles', subs_file, fontsdir=fontfile, force_style=fontstyle),
+      audio, v=1, a=1
+    )
+      .output(filename=output_file)
+      .run(quiet=True, overwrite_output=True)
+  )
+
+
+def str2bool(string):
+    str2val = {"True": True, "False": False}
+    if string in str2val:
+        return str2val[string]
+    else:
+        raise ValueError(
+            f"Expected one of {set(str2val.keys())}, got {string}")
+
+
+def format_timestamp(seconds: float, always_include_hours: bool = False):
+    assert seconds >= 0, "non-negative timestamp expected"
+    milliseconds = round(seconds * 1000.0)
+
+    hours = milliseconds // 3_600_000
+    milliseconds -= hours * 3_600_000
+
+    minutes = milliseconds // 60_000
+    milliseconds -= minutes * 60_000
+
+    seconds = milliseconds // 1_000
+    milliseconds -= seconds * 1_000
+
+    hours_marker = f"{hours}:" if always_include_hours or hours > 0 else ""
+    return f"{hours_marker}{minutes:02d}:{seconds:02d}.{milliseconds:03d}"
+
+
+def write_srt(transcript: Iterator[dict], file: TextIO):
+    for i, segment in enumerate(transcript, start=1):
+        print(
+            f"{i}\n"
+            f"{format_timestamp(segment['start'], always_include_hours=True)} --> "
+            f"{format_timestamp(segment['end'], always_include_hours=True)}\n"
+            f"{segment['text'].strip().replace('-->', '->')}\n",
+            file=file,
+            flush=True,
+        )
+
+
+def filename(path):
+    return os.path.splitext(os.path.basename(path))[0]
+
+
+
+# if __name__ == '__main__':
+#   meta = {
+#     "id": 1576155093245693954,
+#     "ext": 'mp4'
+#   }
+#   tempdirname = Path(f"encoding/temp/{meta['id']}")
+#   video_file_path = f"{meta['id']}.{meta['ext']}"
+#   srt_path = f"{meta['id']}.srt"
+#   out_path = f"{meta['id']}_translated.mp4"
+#   os.chdir(tempdirname)
+#   bake_subs(video_file_path, out_path, srt_path)
+#   anvil_media = anvil.media.from_file(out_path, 'video/mp4')
+#   print(anvil_media)