let5see/AI-Video
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

模板中媒体尺寸改为预置方案

Browse Source
This commit is contained in:
puke
2025-11-12 17:19:06 +08:00
parent 64dcca204e
commit 7443cbf9c2
31 changed files with 576 additions and 90 deletions
Download Patch File Download Diff File Expand all files Collapse all files

7
pixelle_video/pipelines/custom.py
View File

@@ -92,8 +92,7 @@ class CustomPipeline(BasePipeline):
ref_audio: Optional[str] = None,
image_workflow: Optional[str] = None,
image_width: int = 1024,
image_height: int = 1024,
# Note: image_width and image_height are now auto-determined from template
frame_template: Optional[str] = None,
video_fps: int = 30,
@@ -161,6 +160,10 @@ class CustomPipeline(BasePipeline):
generator = HTMLFrameGenerator(template_path)
template_requires_image = generator.requires_image()
# Read media size from template meta tags
image_width, image_height = generator.get_media_size()
logger.info(f"📐 Media size from template: {image_width}x{image_height}")
if template_requires_image:
logger.info(f"📸 Template requires image generation")
else:

16
pixelle_video/pipelines/standard.py
View File

@@ -94,8 +94,7 @@ class StandardPipeline(BasePipeline):
max_image_prompt_words: int = 60,
# === Image Parameters ===
image_width: int = 1024,
image_height: int = 1024,
# Note: image_width and image_height are now auto-determined from template meta tags
image_workflow: Optional[str] = None,
# === Video Parameters ===
@@ -151,9 +150,8 @@ class StandardPipeline(BasePipeline):
min_image_prompt_words: Min image prompt length
max_image_prompt_words: Max image prompt length
image_width: Generated image width (default 1024)
image_height: Generated image height (default 1024)
image_workflow: Image workflow filename (e.g., "image_flux.json", None = use default)
Note: Image/video size is now auto-determined from template meta tags
video_fps: Video frame rate (default 30)
@@ -239,6 +237,16 @@ class StandardPipeline(BasePipeline):
template_config = self.core.config.get("template", {})
frame_template = template_config.get("default_template", "1080x1920/default.html")
# Read media size from template meta tags
from pixelle_video.services.frame_html import HTMLFrameGenerator
from pixelle_video.utils.template_util import resolve_template_path
template_path = resolve_template_path(frame_template)
temp_generator = HTMLFrameGenerator(template_path)
image_width, image_height = temp_generator.get_media_size()
logger.info(f"📐 Media size from template: {image_width}x{image_height}")
# Create storyboard config
config = StoryboardConfig(
task_id=task_id,

52
pixelle_video/services/frame_html.py
View File

@@ -141,6 +141,58 @@ class HTMLFrameGenerator:
logger.debug(f"Template loaded: {len(content)} chars")
return content
def _parse_media_size_from_meta(self) -> tuple[Optional[int], Optional[int]]:
"""
Parse media size from meta tags in template
Looks for meta tags:
- <meta name="template:media-width" content="1024">
- <meta name="template:media-height" content="1024">
Returns:
Tuple of (width, height) or (None, None) if not found
"""
from bs4 import BeautifulSoup
try:
soup = BeautifulSoup(self.template, 'html.parser')
# Find width and height meta tags
width_meta = soup.find('meta', attrs={'name': 'template:media-width'})
height_meta = soup.find('meta', attrs={'name': 'template:media-height'})
if width_meta and height_meta:
width = int(width_meta.get('content', 0))
height = int(height_meta.get('content', 0))
if width > 0 and height > 0:
logger.debug(f"Found media size in meta tags: {width}x{height}")
return width, height
return None, None
except Exception as e:
logger.warning(f"Failed to parse media size from meta tags: {e}")
return None, None
def get_media_size(self) -> tuple[int, int]:
"""
Get media size for image/video generation
Returns media size specified in template meta tags.
Returns:
Tuple of (width, height)
"""
media_width, media_height = self._parse_media_size_from_meta()
if media_width and media_height:
return media_width, media_height
# Fallback to default if not specified (should not happen with properly configured templates)
logger.warning(f"No media size meta tags found in template {self.template_path}, using fallback 1024x1024")
return 1024, 1024
def parse_template_parameters(self) -> Dict[str, Dict[str, Any]]:
"""
Parse custom parameters from HTML template

59
pixelle_video/services/video.py
View File

@@ -224,20 +224,43 @@ class VideoService:
-map "[v]" -map "[a]" output.mp4
"""
try:
inputs = [ffmpeg.input(v) for v in videos]
(
ffmpeg
.concat(*inputs, v=1, a=1)
.output(output)
.overwrite_output()
.run(capture_stdout=True, capture_stderr=True)
# Build filter_complex string manually
n = len(videos)
# Build input stream labels: [0:v][0:a][1:v][1:a]...
stream_spec = "".join([f"[{i}:v][{i}:a]" for i in range(n)])
filter_complex = f"{stream_spec}concat=n={n}:v=1:a=1[v][a]"
# Build ffmpeg command
cmd = ['ffmpeg']
for video in videos:
cmd.extend(['-i', video])
cmd.extend([
'-filter_complex', filter_complex,
'-map', '[v]',
'-map', '[a]',
'-y', # Overwrite output
output
])
# Run command
import subprocess
result = subprocess.run(
cmd,
capture_output=True,
text=True,
check=True
)
logger.success(f"Videos concatenated successfully: {output}")
return output
except ffmpeg.Error as e:
error_msg = e.stderr.decode() if e.stderr else str(e)
except subprocess.CalledProcessError as e:
error_msg = e.stderr if e.stderr else str(e)
logger.error(f"FFmpeg concat filter error: {error_msg}")
raise RuntimeError(f"Failed to concatenate videos: {error_msg}")
except Exception as e:
logger.error(f"Concatenation error: {e}")
raise RuntimeError(f"Failed to concatenate videos: {e}")
def _get_video_duration(self, video: str) -> float:
"""Get video duration in seconds"""
@@ -382,10 +405,17 @@ class VideoService:
# Concatenate original video with black padding
video_stream = ffmpeg.concat(video_stream, black_input.video, v=1, a=0)
# Prepare audio stream
# Prepare audio stream (pad if needed to match target duration)
input_audio = ffmpeg.input(audio)
audio_stream = input_audio.audio.filter('volume', audio_volume)
# Pad audio with silence if video is longer
if video_duration > audio_duration:
pad_duration = video_duration - audio_duration
logger.info(f"Video is longer, padding audio with {pad_duration:.2f}s silence")
# Use apad to add silence at the end
audio_stream = audio_stream.filter('apad', whole_dur=target_duration)
if not video_has_audio:
logger.info(f"Video has no audio stream, adding audio track")
# Video is silent, just add the audio
@@ -398,8 +428,7 @@ class VideoService:
output,
vcodec='libx264', # Re-encode video if padded
acodec='aac',
audio_bitrate='192k',
t=target_duration # Trim to target duration
audio_bitrate='192k'
)
.overwrite_output()
.run(capture_stdout=True, capture_stderr=True)
@@ -426,8 +455,7 @@ class VideoService:
output,
vcodec='libx264', # Re-encode video if padded
acodec='aac',
audio_bitrate='192k',
t=target_duration # Trim to target duration
audio_bitrate='192k'
)
.overwrite_output()
.run(capture_stdout=True, capture_stderr=True)
@@ -452,8 +480,7 @@ class VideoService:
output,
vcodec='libx264', # Re-encode video if padded
acodec='aac',
audio_bitrate='192k',
t=target_duration # Trim to target duration
audio_bitrate='192k'
)
.overwrite_output()
.run(capture_stdout=True, capture_stderr=True)