From d07365daf1e0c5c05c4846341be4ddfdecd6ff28 Mon Sep 17 00:00:00 2001
From: Brandon Thomas <echelon@gmail.com>
Date: Sun, 14 Apr 2024 20:39:18 -0400
Subject: [PATCH] watermark scaling

---
 .../job/job_types/lipsync/sad_talker/process_job.rs   |  5 +++--
 .../job/job_types/workflow/comfy_ui/process_job.rs    |  5 +++--
 .../common_commands/ffmpeg_logo_watermark_command.rs  | 11 +++++++----
 3 files changed, 13 insertions(+), 8 deletions(-)

diff --git a/crates/service/job/inference_job/src/job/job_types/lipsync/sad_talker/process_job.rs b/crates/service/job/inference_job/src/job/job_types/lipsync/sad_talker/process_job.rs
index 0bf3e7ba27..bdeb98fb36 100644
--- a/crates/service/job/inference_job/src/job/job_types/lipsync/sad_talker/process_job.rs
+++ b/crates/service/job/inference_job/src/job/job_types/lipsync/sad_talker/process_job.rs
@@ -27,7 +27,7 @@ use crate::job::job_types::lipsync::sad_talker::resize_image::resize_image;
 use crate::job::job_types::lipsync::sad_talker::sad_talker_inference_command::InferenceArgs;
 use crate::job::job_types::lipsync::sad_talker::validate_job::validate_job;
 use crate::job_dependencies::JobDependencies;
-use crate::util::common_commands::ffmpeg_logo_watermark_command;
+use crate::util::common_commands::ffmpeg_logo_watermark_command::WatermarkArgs;
 
 /// The maximum that either width or height can be
 const MAX_DIMENSION : u32 = 1500;
@@ -254,10 +254,11 @@ pub async fn process_job(args: SadTalkerProcessJobArgs<'_>) -> Result<JobSuccess
 
     let command_exit_status = model_dependencies
         .ffmpeg_watermark_command
-        .execute_inference(ffmpeg_logo_watermark_command::InferenceArgs {
+        .execute_inference(WatermarkArgs {
           video_path: &output_video_fs_path,
           maybe_override_logo_path: None,
           alpha: 0.6,
+          scale: 0.1, // NB: 0.1 is good for the Storyteller logo @ 2653x512 placed on 1024x576 output.
           output_path: &output_video_fs_path_watermark,
         });
 
diff --git a/crates/service/job/inference_job/src/job/job_types/workflow/comfy_ui/process_job.rs b/crates/service/job/inference_job/src/job/job_types/workflow/comfy_ui/process_job.rs
index 8fba46d180..9b73d87ec5 100644
--- a/crates/service/job/inference_job/src/job/job_types/workflow/comfy_ui/process_job.rs
+++ b/crates/service/job/inference_job/src/job/job_types/workflow/comfy_ui/process_job.rs
@@ -40,7 +40,7 @@ use crate::job::job_loop::process_single_job_error::ProcessSingleJobError;
 use crate::job::job_types::workflow::comfy_ui::comfy_ui_inference_command::InferenceArgs;
 use crate::job::job_types::workflow::comfy_ui::validate_job::validate_job;
 use crate::job_dependencies::JobDependencies;
-use crate::util::common_commands::ffmpeg_logo_watermark_command;
+use crate::util::common_commands::ffmpeg_logo_watermark_command::WatermarkArgs;
 
 fn get_file_extension(mimetype: &str) -> Result<&'static str> {
     let ext = match mimetype {
@@ -436,10 +436,11 @@ pub async fn process_job(args: ComfyProcessJobArgs<'_>) -> Result<JobSuccessResu
 
         let command_exit_status = model_dependencies
             .ffmpeg_watermark_command
-            .execute_inference(ffmpeg_logo_watermark_command::InferenceArgs {
+            .execute_inference(WatermarkArgs {
                 video_path: &comfy_output_video_file,
                 maybe_override_logo_path: None,
                 alpha: 0.6,
+                scale: 0.1, // NB: 0.1 is good for the Storyteller logo @ 2653x512 placed on 1024x576 output.
                 output_path: &output_video_fs_path_watermark,
             });
 
diff --git a/crates/service/job/inference_job/src/util/common_commands/ffmpeg_logo_watermark_command.rs b/crates/service/job/inference_job/src/util/common_commands/ffmpeg_logo_watermark_command.rs
index 187a5761ef..68148f8966 100644
--- a/crates/service/job/inference_job/src/util/common_commands/ffmpeg_logo_watermark_command.rs
+++ b/crates/service/job/inference_job/src/util/common_commands/ffmpeg_logo_watermark_command.rs
@@ -47,7 +47,7 @@ pub enum ExecutableOrCommand {
   Command(String),
 }
 
-pub struct InferenceArgs<P: AsRef<Path>> {
+pub struct WatermarkArgs<P: AsRef<Path>> {
   /// -i: video path
   pub video_path: P,
 
@@ -57,6 +57,9 @@ pub struct InferenceArgs<P: AsRef<Path>> {
   /// transparency
   pub alpha: f32,
 
+  /// scale
+  pub scale: f32,
+
   /// --output_audio_filename: output path of wav file.
   pub output_path: P,
 }
@@ -107,7 +110,7 @@ impl FfmpegLogoWatermarkCommand {
 
   pub fn execute_inference<P: AsRef<Path>>(
     &self,
-    args: InferenceArgs<P>,
+    args: WatermarkArgs<P>,
   ) -> CommandExitStatus {
     match self.do_execute_inference(args) {
       Ok(exit_status) => exit_status,
@@ -117,7 +120,7 @@ impl FfmpegLogoWatermarkCommand {
 
   fn do_execute_inference<P: AsRef<Path>>(
     &self,
-    args: InferenceArgs<P>,
+    args: WatermarkArgs<P>,
   ) -> AnyhowResult<CommandExitStatus> {
 
     let mut command = String::new();
@@ -150,7 +153,7 @@ impl FfmpegLogoWatermarkCommand {
     // ffmpeg filter
     command.push_str(" -filter_complex ");
     command.push_str(&format!("\"[1]format=rgba,colorchannelmixer=aa={}[logo];", args.alpha));
-    command.push_str("[logo][0]scale2ref=oh*mdar:ih*0.2[logo][video];");
+    command.push_str(&format!("[logo][0]scale2ref=oh*mdar:ih*{}[logo][video];", args.scale));
     command.push_str("[video][logo]overlay=(main_w-overlay_w)-10:(main_h-overlay_h)-10\"");
     command.push_str(" ");