new video inputs to vst

crates/lib/mysql_queries/src/payloads/generic_inference_args/workflow_payload.rs

@@ -58,6 +58,24 @@ pub struct WorkflowArgs {
     #[serde(skip_serializing_if = "Option::is_none")]
     pub maybe_input_file: Option<MediaFileToken>,
 
+    #[serde(rename = "df")]
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub maybe_input_depth_file: Option<MediaFileToken>,
+
+    #[serde(rename = "nf")]
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub maybe_input_normal_file: Option<MediaFileToken>,
+
+    #[serde(rename = "of")]
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub maybe_input_outline_file: Option<MediaFileToken>,
+
+    // Optional global IP Adapter image media file token
+    // This will apply to the entire workflow / video
+    #[serde(rename = "gi")]
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub global_ip_adapter_token: Option<MediaFileToken>,
+
     #[serde(rename = "out")]
     #[serde(skip_serializing_if = "Option::is_none")]
     pub maybe_output_path: Option<String>,
@@ -127,12 +145,6 @@ pub struct WorkflowArgs {
     #[serde(skip_serializing_if = "Option::is_none")]
     pub travel_prompt: Option<String>,
 
-    // Optional global IP Adapter image media file token
-    // This will apply to the entire workflow / video
-    #[serde(rename = "gi")]
-    #[serde(skip_serializing_if = "Option::is_none")]
-    pub global_ip_adapter_token: Option<MediaFileToken>,
-
     #[deprecated(note = "Use `lipsync_enabled` instead")]
     #[serde(rename = "el")]
     #[serde(skip_serializing_if = "Option::is_none")]

crates/service/web/storyteller_web/src/http_server/endpoints/workflows/enqueue_comfy_ui_handler.rs

@@ -259,6 +259,9 @@ pub async fn enqueue_comfy_ui_handler(
         maybe_json_modifications: request.maybe_json_modifications.clone(),
         maybe_workflow_config: request.maybe_workflow_config.clone(),
         maybe_input_file: request.maybe_input_file.clone(),
+        maybe_input_depth_file: None,
+        maybe_input_normal_file: None,
+        maybe_input_outline_file: None,
         maybe_output_path: request.maybe_output_path.clone(),
         creator_visibility:Some(set_visibility),
         trim_start_seconds: Some(trim_start_seconds),

crates/service/web/storyteller_web/src/http_server/endpoints/workflows/enqueue_video_style_transfer_handler.rs

@@ -72,6 +72,22 @@ pub struct EnqueueVideoStyleTransferRequest {
     /// The input video media file (required)
     input_file: MediaFileToken,
 
+    /// Optional: the depth video file
+    /// The underlying media file must be a video.
+    input_depth_file: Option<MediaFileToken>,
+
+    /// Optional: the normal map video file
+    /// The underlying media file must be a video.
+    input_normal_file: Option<MediaFileToken>,
+
+    /// Optional: the outline video file
+    /// The underlying media file must be a video.
+    input_outline_file: Option<MediaFileToken>,
+
+    /// Optional: Global IP-Adapter image.
+    /// The underlying media file must be an image.
+    global_ipa_media_token: Option<MediaFileToken>,
+
     /// The positive prompt (optional)
     prompt: Option<String>,
     
@@ -120,10 +136,6 @@ pub struct EnqueueVideoStyleTransferRequest {
     /// Optional visibility setting override.
     creator_set_visibility: Option<Visibility>,
 
-    /// Optional Global IP-Adapter image.
-    /// The underlying media file must be an image.
-    global_ipa_media_token: Option<MediaFileToken>,
-
     /// To use prompt traveling, enqueue a prompt with this.
     ///
     /// An example of a prompt travel section:
@@ -393,17 +405,18 @@ pub async fn enqueue_video_style_transfer_handler(
         negative_prompt: request.negative_prompt.new_string_trim_or_empty(),
         travel_prompt: coordinated_args.travel_prompt.new_string_trim_or_empty(),
 
-        // Other inputs
+        // Input files
         maybe_input_file: Some(request.input_file.clone()),
+        maybe_input_depth_file: empty_token_to_null(request.input_depth_file.as_ref()),
+        maybe_input_normal_file: empty_token_to_null(request.input_normal_file.as_ref()),
+        maybe_input_outline_file: empty_token_to_null(request.input_outline_file.as_ref()),
+        global_ip_adapter_token: empty_token_to_null(request.global_ipa_media_token.as_ref()),
+
+        // Other inputs
         style_name: Some(request.style),
         creator_visibility: Some(set_visibility),
         trim_start_milliseconds: Some(trim_start_millis),
         trim_end_milliseconds: Some(trim_end_millis),
-        global_ip_adapter_token: request.global_ipa_media_token
-            .as_ref()
-            // NB: Frontend is mistakenly sending empty string - ignore it
-            .filter(|t| t.as_str().trim_or_empty().is_some())
-            .map(|t| t.clone()),
         strength: maybe_strength,
         frame_skip: request.frame_skip,
 
@@ -486,3 +499,10 @@ pub async fn enqueue_video_style_transfer_handler(
         .content_type("application/json")
         .body(body))
 }
+
+// NB: Frontend is mistakenly sending empty string tokens - ignore those
+fn empty_token_to_null(maybe_token: Option<&MediaFileToken>) -> Option<MediaFileToken> {
+    maybe_token
+        .filter(|t| t.as_str().trim_or_empty().is_some())
+        .map(|t| t.clone())
+}

crates/service/web/storyteller_web/src/http_server/endpoints/workflows/enqueue_workflow_upload_request.rs

@@ -237,6 +237,9 @@ pub async fn enqueue_workflow_upload_request(
         maybe_json_modifications: None,
         maybe_workflow_config: None,
         maybe_input_file: None,
+        maybe_input_depth_file: None,
+        maybe_input_normal_file: None,
+        maybe_input_outline_file: None,
         maybe_output_path: None,
         trim_start_seconds: None,
         trim_end_seconds: None,