image-tab + gallery-build tab; progress fix; speech examples

app.py

@@ -41,7 +41,22 @@ pipe.load_lora_weights(load_file(_lora_path), adapter_name="ingredients")
 pipe.set_adapters("ingredients", LORA_SCALE)
 
 
-def _compose_sheet(paths):
+def _gallery_paths(gallery):
+    paths = []
+    for item in gallery or []:
+        if isinstance(item, (list, tuple)):
+            item = item[0]
+        if isinstance(item, dict):
+            item = item.get("path") or item.get("name") or item.get("image")
+        if isinstance(item, str):
+            paths.append(item)
+    return paths
+
+
+def compose_sheet(gallery):
+    paths = _gallery_paths(gallery)
+    if not paths:
+        raise gr.Error("Upload at least one subject image to build a sheet.")
     imgs = [Image.open(p).convert("RGB") for p in paths]
     if len(imgs) == 1:
         return imgs[0]
@@ -59,13 +74,6 @@ def _compose_sheet(paths):
     return canvas
 
 
-def build_sheet_preview(files):
-    if not files:
-        return None
-    paths = [f if isinstance(f, str) else f.get("path", f.get("name")) for f in files]
-    return _compose_sheet(paths)
-
-
 def _build_prompt(sheet, action):
     return f"Reference sheet: {sheet.strip()}\n\nGenerated video: {action.strip()}"
 
@@ -82,9 +90,9 @@ def _duration(*args, **kwargs):
 
 
 @spaces.GPU(duration=_duration)
-def generate(files, sheet, action, lora_scale, seed, randomize, progress=gr.Progress(track_tqdm=True)):
-    if not files:
-        raise gr.Error("Upload a reference sheet image, or several subject images to build one.")
+def generate(sheet_image, sheet, action, lora_scale, seed, randomize, progress=gr.Progress()):
+    if sheet_image is None:
+        raise gr.Error("Add a reference sheet (upload one, or build one from subject images in the other tab).")
     if not sheet.strip():
         raise gr.Error("Describe the elements in the reference sheet (characters, props, location).")
     if not action.strip():
@@ -94,8 +102,7 @@ def generate(files, sheet, action, lora_scale, seed, randomize, progress=gr.Prog
         seed = random.randint(0, MAX_SEED)
     seed = int(seed)
 
-    paths = [f if isinstance(f, str) else f.get("path", f.get("name")) for f in files]
-    sheet_img = _compose_sheet(paths).resize((WIDTH, HEIGHT), Image.LANCZOS)
+    sheet_img = sheet_image.convert("RGB").resize((WIDTH, HEIGHT), Image.LANCZOS)
     ref = [sheet_img] * NUM_FRAMES
     pipe.set_adapters("ingredients", float(lora_scale))
     prompt = _build_prompt(sheet, action)
@@ -122,21 +129,25 @@ def generate(files, sheet, action, lora_scale, seed, randomize, progress=gr.Prog
 with gr.Blocks(title="LTX-2.3 Ingredients (Fast)") as demo:
     gr.Markdown(
         "# ⚡ LTX-2.3 Ingredients — Fast (Distilled)\n"
-        "Reference-sheet control, fast. Generate a video faithful to a sheet of characters, props and location. "
-        "**Upload a ready sheet, or several subject images and we'll tile them into one.** Using "
+        "Reference-sheet control, fast. Upload a ready sheet, or build one from individual subject images. Using "
         "[LTX 2.3 Distilled](https://huggingface.co/diffusers/LTX-2.3-Distilled-Diffusers) with the "
         "[Ingredients IC-LoRA](https://huggingface.co/ltx-community/LTX-2.3-loras), via diffusers 🧨. "
         "(For maximum fidelity, see the Dev demo.)"
     )
     with gr.Row():
         with gr.Column():
-            files = gr.File(label="Reference sheet (1 image) or subject images (several)",
-                            file_count="multiple", file_types=["image"], type="filepath")
-            sheet_preview = gr.Image(label="Reference sheet used", type="pil", interactive=False)
+            with gr.Tabs():
+                with gr.Tab("Reference sheet"):
+                    sheet_image = gr.Image(type="pil", label="Reference sheet (composite of characters / props / location)")
+                with gr.Tab("Build from subject images"):
+                    gallery = gr.Gallery(label="Upload subject images (characters, props, location)",
+                                         type="filepath", interactive=True, columns=4, height=240)
+                    build_btn = gr.Button("Build reference sheet ➜")
+                    gr.Markdown("*Tiles your images into one sheet and loads it into the **Reference sheet** tab.*")
             sheet = gr.Textbox(label="Reference sheet description", lines=3,
                                placeholder="a young woman with red hair in a green jacket (face close-up + turnaround); a brass pocket watch; a cobblestone alley at night")
-            action = gr.Textbox(label="Generated video — the action / shot, plus any speech & sounds", lines=3,
-                                placeholder="the woman walks down the alley and checks the pocket watch, slow dolly-in; footsteps on cobblestone, a soft voice saying 'almost time', distant city hum")
+            action = gr.Textbox(label="Generated video — action / shot, plus any speech & sounds", lines=3,
+                                placeholder="the woman walks down the alley, checks the pocket watch and whispers 'almost time'; footsteps on cobblestone, distant city hum")
             with gr.Accordion("Settings", open=False):
                 lora_scale = gr.Slider(0.8, 1.8, value=1.4, step=0.05, label="LoRA strength")
                 randomize = gr.Checkbox(True, label="Randomize seed")
@@ -146,25 +157,25 @@ with gr.Blocks(title="LTX-2.3 Ingredients (Fast)") as demo:
             video_out = gr.Video(label="Generated video")
             used_seed = gr.Number(label="Seed used", interactive=False)
 
-    files.change(build_sheet_preview, inputs=files, outputs=sheet_preview)
-    run.click(generate, inputs=[files, sheet, action, lora_scale, seed, randomize], outputs=[video_out, used_seed])
+    build_btn.click(compose_sheet, inputs=gallery, outputs=sheet_image)
+    run.click(generate, inputs=[sheet_image, sheet, action, lora_scale, seed, randomize], outputs=[video_out, used_seed])
 
     gr.Examples(
         examples=[
-            [["examples/sheet_garden.png"],
+            ["examples/sheet_garden.png",
              "a cartoon hedgehog (face close-up and body turnaround) and a cartoon rabbit (turnaround); a green coiled garden hose reel and green spray bottles; the interior of a 'Greenfield Home & Garden' store with shelves of plants",
-             "the hedgehog and the rabbit explore the Greenfield Home & Garden store among the plants and garden tools, the rabbit holding a green spray bottle, warm bright store lighting, playful slow camera; cheerful ambient store sounds and soft footsteps",
+             "the hedgehog waddles up to the camera in the Greenfield Home & Garden store and says cheerfully 'Welcome to Greenfield!', while the rabbit hops past holding a green spray bottle; upbeat store music, the hedgehog's friendly voice and soft footsteps",
              1.4, 42, False],
-            [["examples/sheet_hiker.png"],
+            ["examples/sheet_hiker.png",
              "a young woman hiker in a green shirt and khaki shorts (face close-up and body turnaround); a large blue hiking backpack; a wooden walking stick; a shaggy yak with a colorful woven saddle blanket; a Himalayan stone village with prayer flags and snowy mountains",
-             "the woman loads the blue backpack onto the yak in front of snowy Himalayan peaks and a monastery, gentle handheld camera, soft daylight; wind, distant prayer bells and the yak's low grunt",
+             "the woman loads the blue backpack onto the yak, pats its neck and says warmly 'almost at the summit, buddy', snowy peaks and a monastery behind her; wind, distant prayer bells, her voice and the yak's low grunt",
              1.4, 42, False],
-            [["examples/subj_woman.jpg", "examples/subj_horse.jpg", "examples/subj_landscape.jpg"],
+            ["examples/subj_composite.png",
              "a smiling young woman with curly dark hair; a dappled grey horse; a green misty mountain meadow",
-             "the woman walks up to the grey horse in the misty meadow and gently strokes its neck, soft daylight; gentle wind, a soft horse nicker and distant birdsong",
+             "the woman walks up to the grey horse in the misty meadow, gently strokes its neck and says softly 'good boy, easy now'; a soft horse nicker, her gentle voice, light wind and birdsong",
              1.4, 42, False],
         ],
-        inputs=[files, sheet, action, lora_scale, seed, randomize],
+        inputs=[sheet_image, sheet, action, lora_scale, seed, randomize],
         outputs=[video_out, used_seed], fn=generate, cache_examples=True, cache_mode="lazy",
     )