videopython 0.3.0__py3-none-any.whl → 0.4.1__py3-none-any.whl

videopython/{generation → ai/generation}/image.py

@@ -1,6 +1,3 @@
-import io
-import os
-
 import torch
 from diffusers import DiffusionPipeline
 from PIL import Image

videopython/ai/understanding/transcribe.py

@@ -0,0 +1,37 @@
+from typing import Literal
+
+import whisper
+
+from videopython.base.transcription import Transcription, TranscriptionSegment
+from videopython.base.video import Video
+
+
+class VideoTranscription:
+    def __init__(self, model_name: Literal["tiny", "base", "small", "medium", "large", "turbo"] = "small") -> None:
+        self.model = whisper.load_model(name=model_name)
+
+    def transcribe_video(self, video: Video) -> Transcription:
+        """Transcribes video to text.
+
+        Args:
+            video: Video to transcribe.
+
+        Returns:
+            List of dictionaries with segments of text and their start and end times.
+        """
+        if video.audio.is_silent:
+            return Transcription(segments=[])
+
+        audio = video.audio.to_mono()
+        audio = audio.resample(whisper.audio.SAMPLE_RATE)
+        audio_data = audio.data
+
+        transcription = self.model.transcribe(audio=audio_data, word_timestamps=True)
+
+        transcription_segments = [
+            TranscriptionSegment(start=segment["start"], end=segment["end"], text=segment["text"])
+            for segment in transcription["segments"]
+        ]
+        result = Transcription(segments=transcription_segments)
+
+        return result

videopython/base/combine.py

@@ -0,0 +1,45 @@
+from typing import Literal
+
+import numpy as np
+
+from videopython.base.transforms import ResampleFPS, Resize
+from videopython.base.video import Video
+
+
+class StackVideos:
+    def __init__(self, mode: Literal["horizontal", "vertical"]) -> None:
+        self.mode = mode
+
+    def _validate(self, video1: Video, video2: Video) -> tuple[Video, Video]:
+        video1, video2 = self._align_shapes(video1, video2)
+        video1, video2 = self._align_fps(video1, video2)
+        video1, video2 = self._align_duration(video1, video2)
+        return video1, video2
+
+    def _align_fps(self, video1: Video, video2: Video) -> tuple[Video, Video]:
+        if video1.fps > video2.fps:
+            video1 = ResampleFPS(fps=video2.fps).apply(video1)
+        elif video1.fps < video2.fps:
+            video2 = ResampleFPS(fps=video1.fps).apply(video2)
+        return (video1, video2)
+
+    def _align_shapes(self, video1: Video, video2: Video) -> tuple[Video, Video]:
+        if self.mode == "horizontal":
+            video2 = Resize(height=video1.metadata.height).apply(video2)
+        elif self.mode == "vertical":
+            video2 = Resize(width=video1.metadata.width).apply(video2)
+        return (video1, video2)
+
+    def _align_duration(self, video1: Video, video2: Video) -> tuple[Video, Video]:
+        if len(video1.frames) > len(video2.frames):
+            video1 = video1[: len(video2.frames)]
+        elif len(video1.frames) < len(video2.frames):
+            video2 = video2[: len(video1.frames)]
+        return (video1, video2)
+
+    def apply(self, videos: tuple[Video, Video]) -> Video:
+        videos = self._validate(*videos)
+        axis = 1 if self.mode == "vertical" else 2
+        new_frames = np.concatenate((videos[0].frames, videos[1].frames), axis=axis)
+        new_audio = videos[0].audio.overlay(videos[1].audio)
+        return Video(frames=new_frames, fps=videos[0].fps, audio=new_audio)

videopython/base/effects.py

@@ -156,13 +156,13 @@ class Zoom(Effect):
 
         width = video.metadata.width
         height = video.metadata.height
-        crop_sizes_w, crop_sizes_h = np.linspace(width // self.zoom_factor, width, n_frames), np.linspace(
-            height // self.zoom_factor, height, n_frames
+        crop_sizes_w, crop_sizes_h = (
+            np.linspace(width // self.zoom_factor, width, n_frames),
+            np.linspace(height // self.zoom_factor, height, n_frames),
         )
 
         if self.mode == "in":
             for frame, w, h in tqdm(zip(video.frames, reversed(crop_sizes_w), reversed(crop_sizes_h))):
-
                 x = width / 2 - w / 2
                 y = height / 2 - h / 2
 

videopython/base/transcription.py

@@ -0,0 +1,13 @@
+from dataclasses import dataclass
+
+
+@dataclass
+class TranscriptionSegment:
+    start: float
+    end: float
+    text: str
+
+
+@dataclass
+class Transcription:
+    segments: list[TranscriptionSegment]

videopython/base/transforms.py

@@ -1,7 +1,6 @@
 from abc import ABC, abstractmethod
 from enum import Enum
 from multiprocessing import Pool
-from typing import Literal
 
 import cv2
 import numpy as np
@@ -154,7 +153,6 @@ class CropMode(Enum):
 
 
 class Crop(Transformation):
-
     def __init__(self, width: int, height: int, mode: CropMode = CropMode.CENTER):
         self.width = width
         self.height = height