Update app.py

app.py

@@ -1,11 +1,12 @@
-import os, io, time, base64, random, requests
+import os, io, time, base64, random, subprocess
 from typing import Optional
 from urllib.parse import quote
+
+import requests
 from PIL import Image
 import gradio as gr
-from moviepy.editor import VideoFileClip, concatenate_videoclips
 
-# -------- Modal inference endpoint --------
+# -------- Modal inference endpoint (dev) --------
 INFERENCE_URL = "https://moonmath-ai-dev--moonmath-i2v-backend-moonmathinference-run.modal.run"
 
 # -------- small helpers --------
@@ -45,17 +46,19 @@ def stitch_call(start_img: Image.Image, end_img: Image.Image, prompt: str, seed:
         resp = requests.post(url, files=files, headers=headers, timeout=600)
         ctype = (resp.headers.get("content-type") or "").lower()
 
+        # Raw video bytes
         if "application/json" not in ctype:
             resp.raise_for_status()
             return _save_video_bytes(resp.content, "stitch")
 
+        # JSON with url or base64
         data = resp.json()
-        video_url = data.get("video_url") or data.get("url") or data.get("result") or data.get("output")
-        if isinstance(video_url, str) and video_url.startswith(("http://", "https://")):
+        video_url = data.get("video_url") or data.get("url") or data.get("result")
+        if isinstance(video_url, str) and video_url.startswith("http"):
             b = _download_to_bytes(video_url)
             return _save_video_bytes(b, "stitch")
 
-        video_b64 = data.get("video_b64") or data.get("videoBase64")
+        video_b64 = data.get("video_b64")
         if isinstance(video_b64, str):
             pad = (-len(video_b64)) % 4
             if pad:
@@ -64,78 +67,83 @@ def stitch_call(start_img: Image.Image, end_img: Image.Image, prompt: str, seed:
             return _save_video_bytes(b, "stitch")
 
     except Exception as e:
-        print("stitch_call failed:", e)
+        print("stitch_call error:", e)
 
     return None
 
+# -------- FFmpeg-based concatenation --------
+def concat_videos(vid1: str, vid2: str) -> Optional[str]:
+    if not vid1 or not vid2:
+        return None
+    try:
+        os.makedirs("/tmp", exist_ok=True)
+        out_path = f"/tmp/final_{int(time.time())}.mp4"
+
+        # Create a temporary file list for ffmpeg
+        list_file = f"/tmp/list_{int(time.time())}.txt"
+        with open(list_file, "w") as f:
+            f.write(f"file '{vid1}'\n")
+            f.write(f"file '{vid2}'\n")
+
+        # Run ffmpeg concat
+        subprocess.run(
+            ["ffmpeg", "-y", "-f", "concat", "-safe", "0", "-i", list_file, "-c", "copy", out_path],
+            check=True,
+            stdout=subprocess.PIPE,
+            stderr=subprocess.PIPE,
+        )
+
+        return out_path
+    except Exception as e:
+        print("concat_videos error:", e)
+        return None
+
 # -------- Gradio callbacks --------
 def stitch_12(prompt12, seed, img1, img2):
-    if img1 is None or img2 is None:
-        gr.Warning("Please upload Image 1 and Image 2.")
-        return None
     path = stitch_call(img1, img2, prompt12 or "", int(seed or 0))
-    if path is None:
-        gr.Warning("Stitch 1→2 failed.")
     return path
 
 def stitch_23(prompt23, seed, img2, img3):
-    if img2 is None or img3 is None:
-        gr.Warning("Please upload Image 2 and Image 3.")
-        return None
     path = stitch_call(img2, img3, prompt23 or "", int(seed or 0))
-    if path is None:
-        gr.Warning("Stitch 2→3 failed.")
     return path
 
-def stitch_all(video12, video23):
-    if video12 is None or video23 is None:
-        gr.Warning("Need both videos to stitch all.")
-        return None
-    try:
-        clip1 = VideoFileClip(video12)
-        clip2 = VideoFileClip(video23)
-        final = concatenate_videoclips([clip1, clip2])
-        out_path = f"/tmp/stitch_all_{int(time.time())}.mp4"
-        final.write_videofile(out_path, codec="libx264", audio=False, verbose=False, logger=None)
-        return out_path
-    except Exception as e:
-        print("stitch_all failed:", e)
-        gr.Warning("Stitch All failed.")
+def stitch_all(vid12, vid23):
+    if vid12 is None or vid23 is None:
+        gr.Warning("Generate both videos first before stitching all.")
         return None
+    return concat_videos(vid12, vid23)
 
 # -------- UI --------
 CSS = """
 .gradio-container { padding: 24px; }
-.pill button { border-radius: 999px !important; padding: 10px 18px; }
-.rounded textarea { border-radius: 16px !important; }
 """
 
-with gr.Blocks(css=CSS, title="Stitch Master") as demo:
-    gr.Markdown("## Stitch — Upload 3 images, generate videos between 1→2 and 2→3, then merge them.")
+with gr.Blocks(css=CSS, title="Stitch — 3 uploads, 2 stitches") as demo:
+    gr.Markdown("## Stitch — Upload 3 images → Generate 1→2, 2→3, then combine.")
 
-    # --- Uploads row ---
     with gr.Row():
-        img1 = gr.Image(label="Image 1 upload", type="pil")
-        img2 = gr.Image(label="Image 2 upload", type="pil")
-        img3 = gr.Image(label="Image 3 upload", type="pil")
-
-    seed = gr.Number(value=0, precision=0, label="Seed (0 = random)")
+        # Top: images
+        with gr.Column():
+            img1 = gr.Image(label="Image 1 upload", type="pil")
+            img2 = gr.Image(label="Image 2 upload", type="pil")
+            img3 = gr.Image(label="Image 3 upload", type="pil")
 
-    # --- First stitch ---
-    prompt12 = gr.Textbox(placeholder="Prompt for stitching 1→2", lines=2, label="Prompt (1→2)", elem_classes=["rounded"])
-    btn12 = gr.Button("Stitch 1→2", elem_classes=["pill"])
-    vid12 = gr.Video(label="Video (image 1+2) output", interactive=False)
-
-    # --- Second stitch ---
-    prompt23 = gr.Textbox(placeholder="Prompt for stitching 2→3", lines=2, label="Prompt (2→3)", elem_classes=["rounded"])
-    btn23 = gr.Button("Stitch 2→3", elem_classes=["pill"])
-    vid23 = gr.Video(label="Video (image 2+3) output", interactive=False)
-
-    # --- Final merge ---
-    btn_all = gr.Button("Stitch All", elem_classes=["pill"])
-    vid_all = gr.Video(label="Final concatenated video", interactive=False)
-
-    # --- Wire buttons ---
+    with gr.Row():
+        # Prompts + buttons
+        with gr.Column():
+            seed = gr.Number(value=0, precision=0, label="Seed (0 = random)")
+            prompt12 = gr.Textbox(placeholder="Prompt for stitching 1→2", lines=2, label="Prompt (1→2)")
+            btn12 = gr.Button("Stitch 1&2")
+            prompt23 = gr.Textbox(placeholder="Prompt for stitching 2→3", lines=2, label="Prompt (2→3)")
+            btn23 = gr.Button("Stitch 2&3")
+            btn_all = gr.Button("Stitch All (combine 1→2 and 2→3)")
+
+        with gr.Column():
+            vid12 = gr.Video(label="Video (1→2)")
+            vid23 = gr.Video(label="Video (2→3)")
+            vid_all = gr.Video(label="Final Combined Video")
+
+    # Wire
     btn12.click(stitch_12, inputs=[prompt12, seed, img1, img2], outputs=[vid12])
     btn23.click(stitch_23, inputs=[prompt23, seed, img2, img3], outputs=[vid23])
     btn_all.click(stitch_all, inputs=[vid12, vid23], outputs=[vid_all])