如何用ffmpeg-python实现渐变填充文字并完成音视频混音？现有抠像方案优化咨询

Got it, let's tackle your two main issues step by step—first replacing that clunky keying workflow with a clean alpha mask-based text overlay (no brightness loss!), then adding audio mixing to your final video.

Your original keying approach caused brightness loss because you were double-processing the gradient through color filters. Instead, we'll generate a text alpha mask directly, then use it to "cut out" your gradient image—this preserves the gradient's original brightness perfectly and simplifies the workflow.

Here's the complete code with comments:

import ffmpeg

# Define your file paths
font_path = 'fonts/my_font.ttf'
input_video_path = 'vid01.mp4'
gradient_image_path = 'fill_gradient_gold.png'  # Your pre-made gradient
final_output_path = 'vid01_final.mp4'
music_path = 'background_music.mp3'

# 1. Set up input sources
input_video = ffmpeg.input(input_video_path)
gradient = ffmpeg.input(gradient_image_path)
background_music = ffmpeg.input(music_path).audio

# 2. Generate a text alpha mask
# First, create a black background matching the video's resolution
black_bg = ffmpeg.input(f'color=c=black:s={input_video["w"]}x{input_video["h"]}', format='lavfi')
# Draw white text on the black background
text_mask = black_bg.filter(
    'drawtext',
    text='some text',
    fontcolor='white',
    fontsize=85,
    x='(w-text_w)/2',  # Center horizontally
    y='68-(text_h/2)',  # Your desired vertical position
    fontfile=font_path
)
# Convert the text layer to a grayscale mask (used for alpha transparency)
text_mask = text_mask.filter('format', pix_fmt='gray')

# 3. Apply the mask to your gradient image (creates gradient-filled text)
gradient_text = ffmpeg.filter(
    [gradient, text_mask],
    'overlay',
    x='(W-w)/2', y='(H-h)/2',  # Center gradient on mask
    alpha_mask=1  # Use the text mask as alpha channel
)

# 4. Overlay the gradient text onto your original video
video_with_text = ffmpeg.filter(
    [input_video, gradient_text],
    'overlay',
    x='(W-w)/2', y='(H-h)/2'  # Position text on video
)

# 5. Mix audio: Combine original video audio with background music
mixed_audio = ffmpeg.filter(
    [input_video.audio, background_music],
    'amix',
    inputs=2,
    duration='longest',  # Match longer of the two audio sources
    weights='1 0.5'  # Optional: Adjust volume balance (1 = original audio, 0.5 = music)
)

# 6. Output the final video with text and mixed audio
ffmpeg.output(video_with_text, mixed_audio, final_output_path).overwrite_output().run()

Key Improvements: #

• No brightness loss: We're using your original gradient image directly, only masking out the text shape instead of altering colors via keying.
• Simplified workflow: No multiple splits or colorkey filters—just a single mask generation and overlay step.

• If your original video has no audio, replace the mixed_audio line with mixed_audio = background_music.
• Adjust the weights parameter in amix to tweak volume balance (e.g., weights='1 0.3' makes the music quieter).
• If you want the music to fade out at the end, add an afade filter to the music input:

  background_music = ffmpeg.input(music_path).audio.filter('afade', t='out', st='30', d='5')
  

  (This fades out the music starting at 30 seconds, over 5 seconds.)

内容的提问来源于stack exchange，提问作者Roman March