One tab per reference modality, in reading order

app.py

@@ -375,7 +375,8 @@ load_models()
 INTRO = """# MiniMax-H3 · references
 
 Bring a subject, a style, a camera move or a voice, and generate video with a synchronized soundtrack in one pass.
-References are read **in the order below** — the video first, then the images, then the audio.
+Fill as many tabs as you like — references are read **in tab order**, video then images then audio, and that order is
+part of the request.
 """
 
 with gr.Blocks(title="MiniMax-H3 references") as demo:
@@ -388,11 +389,17 @@ with gr.Blocks(title="MiniMax-H3 references") as demo:
                 lines=3,
                 value="The character walks through a neon-lit street in the rain, humming to themselves",
             )
-            video = gr.Video(label="Motion & camera reference (optional, 2–15 s)")
-            with gr.Row():
-                first_image = gr.Image(label="Reference image", type="filepath")
-                second_image = gr.Image(label="Reference image", type="filepath")
-            audio = gr.Audio(label="Voice or music reference (optional)", type="filepath")
+            # One tab per modality, in the order the model reads them. A reference left in a tab that is not the
+            # open one is still part of the request — the tabs lay the slots out, they do not choose between them.
+            with gr.Tabs():
+                with gr.Tab("Video"):
+                    video = gr.Video(label="Motion & camera, 2–15 s. Its soundtrack comes along.")
+                with gr.Tab("Images"):
+                    with gr.Row():
+                        first_image = gr.Image(label="Subject, style or scene", type="filepath")
+                        second_image = gr.Image(label="Subject, style or scene", type="filepath")
+                with gr.Tab("Audio"):
+                    audio = gr.Audio(label="A voice or a piece of music", type="filepath")
             canvas = gr.Dropdown(label="Canvas", choices=list(CANVASES), value=DEFAULT_CANVAS)
             match = gr.Checkbox(label="Match the reference soundtrack", value=True, visible=False)
             duration = gr.Slider(
@@ -406,7 +413,9 @@ with gr.Blocks(title="MiniMax-H3 references") as demo:
 
     slots = [video, first_image, second_image, audio]
     for control in [*slots, match]:
-        control.change(duration_controls, [*slots, match], [match, duration], show_progress="hidden")
+        control.change(
+            duration_controls, [*slots, match], [match, duration], show_progress="hidden", api_name=False
+        )
 
     run.click(
         generate,