fix cuda acceleration.

Author: fujiawei-dev

README.md

@@ -1,7 +1,7 @@
 <!--
  * @Date: 2021.02.30 17:30:51
  * @LastEditors: Rustle Karl
- * @LastEditTime: 2021.05.04 13:13:00
+ * @LastEditTime: 2021.05.25 08:27:32
 -->
 
 # FFmpeg Generator
@@ -12,6 +12,7 @@ Python bindings for FFmpeg - with almost all filters support, even `gltransition
 
 - [FFmpeg Generator](#ffmpeg-generator)
   - [Overview](#overview)
+  - [TODO](#todo)
   - [Installation](#installation)
   - [Documents](#documents)
   - [GLTransition Filter](#gltransition-filter)
@@ -44,6 +45,11 @@ This project is based on [`ffmpeg-python`](https://github.com/kkroening/ffmpeg-p
 - support FFplay&FFprobe
 - enable cuda hwaccel by default
 
+## TODO
+
+- [] separate outputs for select filter
+- [] experiment: automatically generate subtitle files
+
 ## Installation
 
 ```shell
@@ -64,7 +70,7 @@ Or read my study notes, plan to demonstrate all the filters, but written in Chin
 - [All Examples for Audio Filters](docs/afilters.md)
 - [All Examples for Video Filters](docs/vfilters.md)
 - [All Examples for Audio/Video Sources](docs/sources.md)
-- [All Examples for Media Filters](docs/mfilters.md)
+- [All Examples for Media Filters](docs/avfilters.md)
 - [Introduce Usage of FFplay](docs/ffplay.md)
 - [More Notes](https://github.com/studying-notes/ffmpeg-notes)
 
@@ -264,7 +270,11 @@ vtools.read_frame_as_jpeg(src="src", frame=10)
 ```python
 from ffmpeg import atools
 
-atools.convert_audio_to_raw_pcm(src="src")
+audio = '/path/to/audio.m4a'
+dst = '/path/to/dst.pcm'
+
+atools.convert_audio_to_raw_pcm(src=audio, dst=None)
+atools.convert_audio_to_raw_pcm(src=audio, dst=dst)
 ```
 
 ### Assemble Video from Sequence of Frames

ffmpeg/__init__.py

@@ -2,7 +2,7 @@
 Date: 2021.02.25 14:34:07
 Description: Omit
 LastEditors: Rustle Karl
-LastEditTime: 2021.05.08 08:38:45
+LastEditTime: 2021.05.31 10:38:26
 '''
 import subprocess
 
@@ -48,7 +48,7 @@
     'vtools',
 ]
 
-__version__ = '1.0.1'
+__version__ = '1.0.2'
 
 
 def run_ffmpeg(option: str = None, stdout=None, check=True, **kwargs) -> subprocess.CompletedProcess:

ffmpeg/_ffmpeg.py

@@ -2,7 +2,7 @@
 Date: 2021.02.25 15:02:34
 Description: Omit
 LastEditors: Rustle Karl
-LastEditTime: 2021.05.04 23:37:24
+LastEditTime: 2021.05.10 12:33:20
 '''
 from pathlib import Path
 
@@ -33,12 +33,11 @@ def input(source, video_device: str = None, audio_device: str = None, format: st
 
     kwargs['source'] = str(source)
 
-    if settings.CUDA_ENABLE and enable_cuda and Path(source).suffix not in constants.IMAGE_FORMATS:
+    if settings.CUDA_ENABLE and enable_cuda and \
+            Path(source).suffix not in constants.IMAGE_FORMATS:
+        hwaccel = "cuda"
         if vcodec not in constants.CUDA_ENCODERS:
-            vcodec = None
-
-        kwargs['hwaccel'] = "cuda"
-        kwargs['vcodec'] = settings.DEFAULT_DECODER
+            vcodec = settings.DEFAULT_DECODER
 
     kwargs = drop_empty_dict_values(kwargs, hwaccel=hwaccel, vcodec=vcodec,
                                     f=format, pix_fmt=pixel_format, ss=start_position,
@@ -98,9 +97,12 @@ def output(*streams_or_source, vn=False, an=False, ar=None, ab=None, ac=None,
     if settings.CUDA_ENABLE and enable_cuda and not preview and \
             Path(kwargs['source']).suffix not in constants.IMAGE_FORMATS:
         if vcodec not in constants.CUDA_ENCODERS:
-            vcodec = None
+            vcodec = settings.DEFAULT_ENCODER
 
-        kwargs['vcodec'] = settings.DEFAULT_ENCODER
+    # codec over acodec/vcodec
+    if codec is not None:
+        acodec = None
+        vcodec = None
 
     if video_bitrate is not None:
         kwargs['b:v'] = video_bitrate

ffmpeg/_node.py

@@ -2,10 +2,11 @@
 Date: 2021.02.25 14:34:07
 Description: Omit
 LastEditors: Rustle Karl
-LastEditTime: 2021.04.29 11:01:29
+LastEditTime: 2021.05.24 07:34:34
 '''
 from __future__ import annotations
 
+from enum import Enum
 from typing import Any, Dict, List, Tuple, Type, Union
 
 from ._dag import DagEdge, DagNode, Edge, get_outgoing_edges
@@ -114,7 +115,9 @@ def __getitem__(self, index: str) -> Stream:
         return self._node.stream(label=self._label, selector=index)
 
 
-class NodeTypes(object):
+class NodeTypes(str, Enum):
+    # https://stackoverflow.com/questions/58608361/string-based-enum-in-python
+
     Base = 'Base'
     Input = 'Input'
     Filter = 'Filter'

ffmpeg/_utils.py

@@ -4,7 +4,7 @@
 LastEditors: Rustle Karl
 LastEditTime: 2021.05.04 23:37:10
 '''
-from typing import Dict, Iterable, List
+from typing import Dict, Iterable, List, Union
 
 _backslash = '\\'
 _empty_symbols = (None, '', [], {})  # exclude 0
@@ -83,11 +83,20 @@ def string_to_seconds(clock: str) -> int:
     if isinstance(clock, (int, float)):
         return clock
 
-    hours, minutes, seconds = [int(c) for c in clock.split(":")]
+    clock = [int(c) for c in clock.split(":")]
+    if len(clock) == 0:
+        hours, minutes, seconds = 0, 0, 0
+    elif len(clock) == 1:
+        hours, minutes, seconds = 0, 0, clock[0]
+    elif len(clock) == 2:
+        hours, minutes, seconds = 0, clock[0], clock[1]
+    else:
+        hours, minutes, seconds = clock[0], clock[1], clock[2]
+
     return hours * 60 * 60 + minutes * 60 + seconds
 
 
-def seconds_to_string(seconds: float) -> str:
+def seconds_to_string(seconds: Union[float, int, str]) -> str:
     if isinstance(seconds, str):
         return seconds
 

ffmpeg/nodes.py

@@ -8,6 +8,7 @@
 
 import copy
 import os
+import shutil
 import subprocess
 from collections import defaultdict
 from pathlib import Path
@@ -119,6 +120,11 @@ def run_async(self, *, executable="ffmpeg", direct_print=True, join_args=False,
                   pipe_stdin=False, pipe_stdout=True, pipe_stderr=True, quiet=False,
                   overwrite=True, progress='') -> subprocess.Popen:
         '''Asynchronously invoke ffmpeg for the supplied node graph.'''
+        if shutil.which(executable) is None:
+            raise FileNotFoundError(f"Can't find {executable} in $PATH or "
+                                    f"current directory. Please specify a absolute path or "
+                                    f"add {executable} into $PATH.")
+
         cmd_args_seq = self.compile(
                 executable=executable,
                 direct_print=direct_print,
@@ -179,9 +185,10 @@ def output(self, *streams_or_source, vn=False, an=False, ar=None, ab=None, ac=No
                acodec=None, vcodec=None, codec: str = None, aq_scale=None, vq_scale=None,
                aspect=None, fps=None, format=None, pixel_format=None, video_bitrate=None,
                audio_bitrate=None, v_profile=None, preset=None, mov_flags=None,
-               shortest=False, frame_size=None, v_frames: int = None, start_position: float = None,
-               duration: float = None, video_filter: str = None, audio_filter: str = None, ignore_output=False,
-               preview: bool = False, enable_cuda=True, args: list = None, **kwargs) -> OutputStream:
+               shortest=False, frame_size=None, v_frames: int = None, duration: Union[float, int, str] = None,
+               start_position: Union[float, int, str] = None, video_filter: str = None,
+               audio_filter: str = None, ignore_output=False, preview: bool = False,
+               enable_cuda=True, args: list = None, **kwargs) -> OutputStream:
         raise NotImplementedError
 
     def filter(self, *args, **kwargs) -> FilterableStream:

requirements.txt

@@ -1,8 +1 @@
-graphviz~=0.16
-matplotlib~=3.2.2
-numpy~=1.19.5
-pillow~=8.0.1
 project-pkgs~=1.0.0
-setuptools~=49.2.1
-tqdm~=4.47.0
-psutil~=5.8.0

setup.py

@@ -11,9 +11,7 @@
 
 # What packages are required for this module to be executed?
 requires = [
-    'graphviz',
     'project-pkgs',
-    'tqdm',
 ]
 
 # Import the README and use it as the long-description.