Update app.py

app.py

@@ -274,8 +274,10 @@ def _collect_character_refs(reference_files, reference_zip) -> list[CharacterRef
 
 
 def _text_args(text_prompt: str) -> list[str]:
-    words = [part.strip() for part in (text_prompt or "").split() if part.strip()]
-    return words or ["human", "character"]
+    if not text_prompt or not text_prompt.strip():
+        return ["human character"]
+    parts = [part.strip() for part in re.split(r"[\n,]+", text_prompt) if part.strip()]
+    return parts or ["human character"]
 
 
 def _run_command(command: list[str], progress=None) -> str:
@@ -742,7 +744,11 @@ def build_ui():
                 simple_ref = gr.Image(type="filepath", label="Reference image")
                 simple_driving = gr.Video(label="Driving / source video")
             simple_prompt = gr.Textbox(label="Prompt for SCAIL-2", lines=3)
-            simple_sam3_text = gr.Textbox(value="human character", label="SAM3 text prompt")
+            simple_sam3_text = gr.Textbox(
+                value="human character",
+                label="SAM3 text prompt",
+                info="Use one or more comma/newline-separated prompts, e.g. `human character` or `woman in red dress, man in blue jacket`.",
+            )
             with gr.Accordion("Animation options", open=True):
                 simple_max_persons = gr.Number(value=2, precision=0, label="Max tracked subjects")
                 simple_crop = gr.Dropdown(
@@ -789,6 +795,12 @@ def build_ui():
             )
             with gr.Accordion("Reference naming", open=True):
                 gr.Markdown(
+                    "The Advanced Character Pack uses a simple identity convention: "
+                    "`character_0` must match the left-most tracked subject in the driving video, "
+                    "`character_1` the next subject to the right, and so on. "
+                    "Name your reference files according to that left-to-right order.\n\n"
+                    "Example: if the driving video starts with the woman on the left and the man on "
+                    "the right, use `character_0` for the woman and `character_1` for the man.\n\n"
                     "Upload reference images with names like:\n\n"
                     "```text\n"
                     "character_0__front.png\n"
@@ -805,8 +817,9 @@ def build_ui():
                     "  character_1/\n"
                     "    front.png\n"
                     "```\n\n"
-                    "Character colors are assigned by slot order: `character_0`, then `character_1`, etc. "
-                    "The driving video is tracked left-to-right with the same color order."
+                    "All views of the same character should use the same `character_N` prefix. "
+                    "If the detected order is ambiguous or characters cross over, use a simple case first; "
+                    "manual track assignment is not part of this V1."
                 )
             with gr.Row():
                 adv_refs = gr.Files(
@@ -822,7 +835,11 @@ def build_ui():
             adv_driving = gr.Video(label="Driving video")
             adv_prompt = gr.Textbox(label="Prompt for SCAIL-2", lines=3)
             with gr.Row():
-                adv_sam3_text = gr.Textbox(value="human character", label="SAM3 text prompt")
+                adv_sam3_text = gr.Textbox(
+                    value="human character",
+                    label="SAM3 text prompt",
+                    info="Use broad prompts to detect all subjects, or comma/newline-separated prompts for distinct subjects.",
+                )
                 adv_max_subjects = gr.Number(value=2, precision=0, label="Max tracked subjects")
             adv_run = gr.Button("Generate advanced character pack", variant="primary")
             adv_gallery = gr.Gallery(
@@ -896,4 +913,4 @@ def build_ui():
 
 
 if __name__ == "__main__":
-    build_ui().queue(max_size=4).launch(show_error=True)
+    build_ui().queue(max_size=4).launch(show_error=True)