Update app.py

app.py

@@ -1,6 +1,6 @@
 import os, io, time, base64, random, subprocess
 from typing import Optional, List
-from urllib.parse import quote
+from urllib.parse import urlencode
 
 import requests
 from PIL import Image
@@ -30,17 +30,42 @@ def _download_to_bytes(url: str) -> bytes:
     r.raise_for_status()
     return r.content
 
-def stitch_call(start_img: Image.Image, end_img: Image.Image, prompt: str, seed: Optional[int]) -> Optional[str]:
+def stitch_call(
+    start_img: Image.Image,
+    end_img: Image.Image,
+    prompt: str,
+    seed: Optional[int],
+    negative_prompt: Optional[str] = None,
+    frames_per_second: int = 24,
+    video_length: int = 4,
+    num_inference_steps: Optional[int] = None,
+) -> Optional[str]:
     """
-    Calls your Modal backend with two images + prompt + seed and returns a local /tmp video path.
+    Required (in body): image_bytes (+ image_bytes_end)
+    In URL query: prompt, negative_prompt, frames_per_second, video_length, seed, num_inference_steps
     """
     if start_img is None or end_img is None:
         return None
 
+    # default seed behavior
     if seed in (None, 0, -1):
         seed = random.randint(1, 2**31 - 1)
 
-    url = f"{INFERENCE_URL}?prompt={quote(prompt or '')}&seed={seed}"
+    # Build query string
+    q = {
+        "prompt": prompt or "",
+        "seed": int(seed),
+        "frames_per_second": int(frames_per_second),
+        "video_length": int(video_length),
+    }
+    if negative_prompt:
+        q["negative_prompt"] = negative_prompt
+    if num_inference_steps is not None:
+        q["num_inference_steps"] = int(num_inference_steps)
+
+    url = f"{INFERENCE_URL}?{urlencode(q)}"
+
+    # Images go in the body
     files = {
         "image_bytes": ("start.png", _png_bytes(start_img), "image/png"),
         "image_bytes_end": ("end.png", _png_bytes(end_img), "image/png"),
@@ -135,8 +160,10 @@ def collect_choices(*imgs):
             choices.append(str(i))
     return gr.update(choices=choices), gr.update(choices=choices)
 
-def stitch_selected(prompt, seed, start_idx_str, end_idx_str, *imgs):
-    """Run inference for selected start/end indices (1-based strings)."""
+def stitch_selected(
+    prompt, negative_prompt, fps, length_sec, seed, start_idx_str, end_idx_str, *imgs
+):
+    """Run inference for selected start/end indices (1-based strings) + options."""
     if not start_idx_str or not end_idx_str:
         gr.Warning("Please select Start and End frames.")
         return None
@@ -157,7 +184,19 @@ def stitch_selected(prompt, seed, start_idx_str, end_idx_str, *imgs):
         gr.Warning("Selected slots are empty.")
         return None
 
-    vid = stitch_call(start_img, end_img, prompt or "", int(seed or 0))
+    fps_val = int(str(fps)) if fps else 24
+    len_val = int(str(length_sec)) if length_sec else 4
+
+    vid = stitch_call(
+        start_img=start_img,
+        end_img=end_img,
+        prompt=prompt or "",
+        seed=int(seed or 0),
+        negative_prompt=(negative_prompt or "").strip() or None,
+        frames_per_second=fps_val,
+        video_length=len_val,
+        num_inference_steps=None,
+    )
     if not vid:
         gr.Warning("Generation failed.")
         return None
@@ -246,7 +285,7 @@ with gr.Blocks(css=CSS, title="StitchTool") as demo:
         outputs=img_comps
     )
 
-    # Seed + Start/End selection + Prompt + Stitch + Preview
+    # Seed + Start/End selection + Prompt + options + Stitch + Preview
     seed = gr.Number(value=0, precision=0, label="Seed (0 = random)")
 
     with gr.Row():
@@ -254,12 +293,35 @@ with gr.Blocks(css=CSS, title="StitchTool") as demo:
         with gr.Column(scale=1, min_width=420, elem_classes=["stitch-box"]):
             start_dd = gr.Dropdown(label="Start frame", choices=[], interactive=True)
             end_dd = gr.Dropdown(label="End frame", choices=[], interactive=True)
+
             prompt = gr.Textbox(
                 placeholder="Describe the transition between the selected start and end frames…",
                 lines=3,
                 label="Prompt",
                 elem_classes=["rounded"]
             )
+
+            negative = gr.Textbox(
+                placeholder="Optional: things to avoid (e.g., 'no cuts, no angle switch, no text overlays')",
+                lines=2,
+                label="Negative prompt",
+                elem_classes=["rounded"]
+            )
+
+            with gr.Row():
+                fps = gr.Dropdown(
+                    label="Frame rate",
+                    choices=["16", "24", "32"],
+                    value="24",
+                    interactive=True,
+                )
+                length_sec = gr.Dropdown(
+                    label="Video length (sec)",
+                    choices=["2", "4"],
+                    value="4",
+                    interactive=True,
+                )
+
             run_btn = gr.Button("Generate", elem_classes=["pill"])
             add_tl_btn = gr.Button("Add to timeline", elem_classes=["pill"])
 
@@ -278,7 +340,7 @@ with gr.Blocks(css=CSS, title="StitchTool") as demo:
     # stitch action → preview
     run_btn.click(
         fn=stitch_selected,
-        inputs=[prompt, seed, start_dd, end_dd] + img_comps,
+        inputs=[prompt, negative, fps, length_sec, seed, start_dd, end_dd] + img_comps,
         outputs=[preview]
     )
 
@@ -306,4 +368,4 @@ with gr.Blocks(css=CSS, title="StitchTool") as demo:
     )
 
 if __name__ == "__main__":
-    demo.queue().launch()
+    demo.queue().launch()