Merge pull request #388 from philomena-dev/transparency-fixes

Fix transparency issues in webm videos
derpibooru · Dec 26, 2024 · 0a48359 · 0a48359
2 parents d695bf1 + 9b022e7
commit 0a48359
Show file tree

Hide file tree

Showing 2 changed files with 68 additions and 17 deletions.
diff --git a/lib/philomena_media/gif_preview.ex b/lib/philomena_media/gif_preview.ex
@@ -13,6 +13,9 @@ defmodule PhilomenaMedia.GifPreview do
           target_framerate: target_framerate()
         ]
 
+  @typedoc "One of av1, h264, libvpx, libvpx-vp9"
+  @type decoder :: String.t()
+
   @doc """
   Generate a GIF preview of the given video input with evenly-spaced sample points.
 
@@ -31,8 +34,8 @@ defmodule PhilomenaMedia.GifPreview do
     * 1 or above: 5 images
     * otherwise: 2 images
   """
-  @spec preview(Path.t(), Path.t(), duration(), dimensions(), opts()) :: :ok
-  def preview(video, gif, duration, dimensions, opts \\ []) do
+  @spec preview(decoder(), Path.t(), Path.t(), duration(), dimensions(), opts()) :: :ok
+  def preview(decoder, video, gif, duration, dimensions, opts \\ []) do
     target_framerate = Keyword.get(opts, :target_framerate, 2)
 
     num_images =
@@ -48,22 +51,41 @@ defmodule PhilomenaMedia.GifPreview do
     {_output, 0} =
       System.cmd(
         "ffmpeg",
-        commands(video, gif, clamp(duration), dimensions, num_images, target_framerate)
+        commands(decoder, video, gif, clamp(duration), dimensions, num_images, target_framerate)
       )
 
     :ok
   end
 
-  @spec commands(Path.t(), Path.t(), duration(), dimensions(), num_images(), target_framerate()) ::
+  @spec commands(
+          decoder(),
+          Path.t(),
+          Path.t(),
+          duration(),
+          dimensions(),
+          num_images(),
+          target_framerate()
+        ) ::
           [String.t()]
-  defp commands(video, gif, duration, {target_width, target_height}, num_images, target_framerate) do
+  defp commands(
+         decoder,
+         video,
+         gif,
+         duration,
+         {target_width, target_height},
+         num_images,
+         target_framerate
+       ) do
     # Compute range [0, num_images)
     image_range = 0..(num_images - 1)
 
     # Generate input list in the following form:
-    #   -ss 0.0 -i input.webm
+    #   -ss 0.0 -c:v libvpx -i input.webm
     input_arguments =
-      Enum.flat_map(image_range, &["-ss", "#{&1 * duration / num_images}", "-i", video])
+      Enum.flat_map(
+        image_range,
+        &["-ss", "#{&1 * duration / num_images}", "-c:v", decoder, "-i", video]
+      )
 
     # Generate graph in the following form:
     #   [0:v] trim=end_frame=1 [t0]; [1:v] trim=end_frame=1 [t1] ...
@@ -87,7 +109,7 @@ defmodule PhilomenaMedia.GifPreview do
       "[s0] palettegen=stats_mode=single:max_colors=255:reserve_transparent=1 [palettegen]"
 
     paletteuse_filter =
-      "[s1][palettegen] paletteuse=dither=bayer:bayer_scale=5:new=1:alpha_threshold=255"
+      "[s1][palettegen] paletteuse=dither=bayer:bayer_scale=5:new=1:alpha_threshold=251"
 
     filter_graph =
       [

diff --git a/lib/philomena_media/processors/webm.ex b/lib/philomena_media/processors/webm.ex
@@ -29,11 +29,12 @@ defmodule PhilomenaMedia.Processors.Webm do
     duration = analysis.duration
     stripped = strip(file)
     preview = preview(duration, stripped)
-    mp4 = scale_mp4_only(stripped, dimensions, dimensions)
+    decoder = select_decoder(file)
+    mp4 = scale_mp4_only(decoder, stripped, dimensions, dimensions)
 
     {:ok, intensities} = Intensities.file(preview)
 
-    scaled = Enum.flat_map(versions, &scale(stripped, duration, dimensions, &1))
+    scaled = Enum.flat_map(versions, &scale(decoder, stripped, duration, dimensions, &1))
     mp4 = [{:copy, mp4, "full.mp4"}]
 
     [
@@ -82,12 +83,12 @@ defmodule PhilomenaMedia.Processors.Webm do
     stripped
   end
 
-  defp scale(file, duration, dimensions, {thumb_name, target_dimensions}) do
-    {webm, mp4} = scale_videos(file, dimensions, target_dimensions)
+  defp scale(decoder, file, duration, dimensions, {thumb_name, target_dimensions}) do
+    {webm, mp4} = scale_videos(decoder, file, dimensions, target_dimensions)
 
     cond do
       thumb_name in [:thumb, :thumb_small, :thumb_tiny] ->
-        gif = scale_gif(file, duration, dimensions, target_dimensions)
+        gif = scale_gif(decoder, file, duration, dimensions, target_dimensions)
 
         [
           {:copy, webm, "#{thumb_name}.webm"},
@@ -103,7 +104,7 @@ defmodule PhilomenaMedia.Processors.Webm do
     end
   end
 
-  defp scale_videos(file, dimensions, target_dimensions) do
+  defp scale_videos(decoder, file, dimensions, target_dimensions) do
     filter = scale_filter(dimensions, target_dimensions)
     webm = Briefly.create!(extname: ".webm")
     mp4 = Briefly.create!(extname: ".mp4")
@@ -113,6 +114,8 @@ defmodule PhilomenaMedia.Processors.Webm do
         "-loglevel",
         "0",
         "-y",
+        "-c:v",
+        decoder,
         "-i",
         file,
         "-c:v",
@@ -162,7 +165,7 @@ defmodule PhilomenaMedia.Processors.Webm do
     {webm, mp4}
   end
 
-  defp scale_mp4_only(file, dimensions, target_dimensions) do
+  defp scale_mp4_only(decoder, file, dimensions, target_dimensions) do
     filter = scale_filter(dimensions, target_dimensions)
     mp4 = Briefly.create!(extname: ".mp4")
 
@@ -171,6 +174,8 @@ defmodule PhilomenaMedia.Processors.Webm do
         "-loglevel",
         "0",
         "-y",
+        "-c:v",
+        decoder,
         "-i",
         file,
         "-c:v",
@@ -197,15 +202,39 @@ defmodule PhilomenaMedia.Processors.Webm do
     mp4
   end
 
-  defp scale_gif(file, duration, dimensions, target_dimensions) do
+  defp scale_gif(decoder, file, duration, dimensions, target_dimensions) do
     {width, height} = box_dimensions(dimensions, target_dimensions)
     gif = Briefly.create!(extname: ".gif")
 
-    GifPreview.preview(file, gif, duration, {width, height})
+    GifPreview.preview(decoder, file, gif, duration, {width, height})
 
     gif
   end
 
+  defp select_decoder(file) do
+    {output, 0} =
+      System.cmd("ffprobe", [
+        "-loglevel",
+        "0",
+        "-select_streams",
+        "v:0",
+        "-show_entries",
+        "stream=codec_name",
+        "-of",
+        "default=noprint_wrappers=1:nokey=1",
+        "-i",
+        file
+      ])
+
+    # Mediatools verifies that we only have one video stream and that it is
+    # one of the supported formats, so the following is safe to do:
+    case output do
+      "vp8\n" -> "libvpx"
+      "vp9\n" -> "libvpx-vp9"
+      "av1\n" -> "av1"
+    end
+  end
+
   defp scale_filter(dimensions, target_dimensions) do
     {width, height} = box_dimensions(dimensions, target_dimensions)
     "scale=w=#{width}:h=#{height},setsar=1"