Update app.py

app.py

@@ -1,4 +1,4 @@
-import os, json, tempfile, subprocess, shutil, time, uuid
+import os, json, tempfile, subprocess, shutil, uuid
 from pathlib import Path
 from typing import Optional, Tuple, List
 
@@ -15,12 +15,14 @@ OUT_DIR = ROOT / "outputs"
 ASSETS = ROOT / "assets"
 ASSETS.mkdir(exist_ok=True)
 
-BILS_BRAND = os.environ.get("BILS_BRAND", "Bilsimaging · Foley Studio")
-PRIMARY_COLOR = os.environ.get("PRIMARY_COLOR", "#6B5BFF")  # UI accent
+# You can keep these env vars silently; we just won't mention them in the UI
+APP_TITLE   = os.environ.get("APP_TITLE", "Foley Studio · ZeroGPU")
+APP_TAGLINE = os.environ.get("APP_TAGLINE", "Generate scene-true foley for short clips (ZeroGPU-ready).")
+PRIMARY_COLOR = os.environ.get("PRIMARY_COLOR", "#6B5BFF")
 
-MAX_SECS = int(os.environ.get("MAX_SECS", "22"))  # ZeroGPU-friendly clip length
+MAX_SECS = int(os.environ.get("MAX_SECS", "22"))   # ZeroGPU-friendly clip length
 TARGET_H = int(os.environ.get("TARGET_H", "480"))  # downscale target height
-SR = int(os.environ.get("TARGET_SR", "48000"))     # output WAV sample rate
+SR       = int(os.environ.get("TARGET_SR", "48000"))  # WAV sample rate
 
 def sh(cmd: str):
     print(">>", cmd)
@@ -38,20 +40,19 @@ def ffprobe_duration(path: str) -> float:
 
 def _clone_without_lfs():
     """
-    Try a shallow clone while skipping LFS smudge so demo MP4s aren't pulled.
-    Falls back to sparse checkout with needed paths only.
+    Clone repo while skipping LFS smudge to avoid demo video downloads.
+    Falls back to sparse checkout with only essential paths.
     """
     if REPO_DIR.exists():
         return
 
-    # Attempt 1: shallow clone, disable LFS filters
+    # Attempt 1: shallow clone with LFS disabled
     try:
         sh(
             "GIT_LFS_SKIP_SMUDGE=1 "
             "git -c filter.lfs.smudge= -c filter.lfs.required=false "
             f"clone --depth 1 https://github.com/Tencent-Hunyuan/HunyuanVideo-Foley.git {REPO_DIR}"
         )
-        # Optional: remove assets folder if present
         assets = REPO_DIR / "assets"
         if assets.exists():
             shutil.rmtree(assets, ignore_errors=True)
@@ -67,8 +68,6 @@ def _clone_without_lfs():
         "remote add origin https://github.com/Tencent-Hunyuan/HunyuanVideo-Foley.git"
     )
     sh(f"git -C {REPO_DIR} config core.sparseCheckout true")
-
-    # Choose only essential paths
     sparse_file = REPO_DIR / ".git" / "info" / "sparse-checkout"
     sparse_file.parent.mkdir(parents=True, exist_ok=True)
     sparse_file.write_text("\n".join([
@@ -79,8 +78,7 @@ def _clone_without_lfs():
         "LICENSE",
         "README.md",
     ]) + "\n")
-
-    # Branch might be main; change to master if needed
+    # Try main, fallback to master
     try:
         sh(f"git -C {REPO_DIR} fetch --depth 1 origin main")
         sh(f"git -C {REPO_DIR} checkout main")
@@ -89,9 +87,8 @@ def _clone_without_lfs():
         sh(f"git -C {REPO_DIR} checkout master")
 
 def prepare_once():
-    """Clone code (skipping LFS), download weights, set env, prepare dirs."""
+    """Clone code (skip LFS), download weights, set env, prepare dirs."""
     _clone_without_lfs()
-
     WEIGHTS_DIR.mkdir(parents=True, exist_ok=True)
     snapshot_download(
         repo_id="tencent/HunyuanVideo-Foley",
@@ -100,7 +97,6 @@ def prepare_once():
         repo_type="model",
     )
     os.environ["HIFI_FOLEY_MODEL_PATH"] = str(WEIGHTS_DIR)
-
     CACHE_DIR.mkdir(exist_ok=True)
     OUT_DIR.mkdir(exist_ok=True)
 
@@ -120,7 +116,6 @@ def preprocess_video(in_path: str) -> Tuple[str, float]:
     temp_dir = Path(tempfile.mkdtemp(prefix="pre_"))
     trimmed = temp_dir / "trim.mp4"
     processed = temp_dir / "proc.mp4"
-
     trim_args = ["-t", str(MAX_SECS)] if dur > MAX_SECS else []
 
     # Normalize container & remove audio
@@ -133,7 +128,7 @@ def preprocess_video(in_path: str) -> Tuple[str, float]:
         f"\"{trimmed}\""
     ]))
 
-    # Downscale to TARGET_H; ensure mod2 width, baseline profile for compatibility
+    # Downscale to TARGET_H; ensure mod2 width, baseline profile
     vf = f"scale=-2:{TARGET_H}:flags=bicubic"
     sh(" ".join([
         "ffmpeg", "-y", "-i", f"\"{trimmed}\"",
@@ -150,7 +145,7 @@ def preprocess_video(in_path: str) -> Tuple[str, float]:
     return str(processed), final_dur
 
 # ========= Inference (ZeroGPU) =========
-@spaces.GPU(duration=240)  # ~4 minutes per call (fits ZeroGPU window)
+@spaces.GPU(duration=240)  # ~4 minutes per call (ZeroGPU window)
 def run_model(video_path: str, prompt_text: str) -> str:
     """
     Call Tencent's infer.py on GPU and return a 48 kHz WAV path.
@@ -207,16 +202,16 @@ def single_generate(video: str, prompt: str, want_mux: bool, project_name: str)
     try:
         if not video:
             return None, None, "⚠️ Please upload a video.", history
-        history.append(["Preprocess", "Downscaling / trimming…"])
+        history.append(["Preprocess", "Downscaling & trimming"])
         pre_path, final_dur = preprocess_video(video)
-        history.append(["Inference", "Generating foley on GPU…"])
+        history.append(["Inference", "Running on ZeroGPU"])
         wav = run_model(pre_path, prompt or "")
         muxed = None
         if want_mux:
-            history.append(["Mux", "Combining foley with video…"])
+            history.append(["Mux", "Merging foley with video"])
             muxed = mux_audio_with_video(pre_path, wav)
-        history.append(["Done", f"OK · Duration ~{final_dur:.1f}s"])
-        return wav, muxed, f"✅ Finished (≈ {final_dur:.1f}s)", history
+        history.append(["Done", f"OK · ~{final_dur:.1f}s"])
+        return wav, muxed, f"✅ Completed (~{final_dur:.1f}s)", history
     except Exception as e:
         history.append(["Error", str(e)])
         return None, None, f"❌ {type(e).__name__}: {e}", history
@@ -234,53 +229,107 @@ def batch_lite_generate(files: List[str], prompt: str, want_mux: bool) -> Tuple[
         try:
             log.append([f"Preprocess {i}", Path(f).name])
             pre, final_dur = preprocess_video(f)
-            log.append([f"Run {i}", f"GPU infer ~{final_dur:.1f}s"])
+            log.append([f"Run {i}", f"ZeroGPU ~{final_dur:.1f}s"])
             wav = run_model(pre, prompt or "")
             muxed = mux_audio_with_video(pre, wav) if want_mux else None
             outputs.append((wav, muxed))
             log.append([f"Done {i}", "OK"])
         except Exception as e:
             log.append([f"Error {i}", str(e)])
+
     manifest = OUT_DIR / f"batchlite_{uuid.uuid4().hex[:6]}.json"
     manifest.write_text(json.dumps(
         [{"wav": w, "video": v} for (w, v) in outputs], ensure_ascii=False, indent=2
     ))
     return f"✅ Batch-lite finished · items: {len(outputs)}", log
 
-# ========= UI =========
+# ========= UI (refreshed design) =========
 THEME_CSS = f"""
 :root {{
   --brand: {PRIMARY_COLOR};
+  --bg: #0f1120;
+  --panel: #181a2e;
+  --text: #edf0ff;
+  --muted: #b7bce3;
+  --card: #15172a;
 }}
 .gradio-container {{
-  font-family: ui-sans-serif, system-ui, -apple-system, Segoe UI, Roboto, Cairo, Noto Sans, Arial, "Apple Color Emoji", "Segoe UI Emoji";
+  font-family: Inter, ui-sans-serif, -apple-system, Segoe UI, Roboto, Cairo, Noto Sans, Arial;
+  background: var(--bg);
+  color: var(--text);
+}}
+#hero {{
+  background: linear-gradient(135deg, var(--brand) 0%, #2f2e8b 40%, #1b1a3a 100%);
+  border-radius: 18px;
+  padding: 18px 20px;
+  color: white;
+  box-shadow: 0 10px 30px rgba(0,0,0,.35);
 }}
-#brandbar {{
-  background: linear-gradient(90deg, var(--brand), #222);
-  color: white; padding: 12px 16px; border-radius: 12px;
+#hero h1 {{
+  margin: 0 0 6px 0;
+  font-size: 20px;
+  font-weight: 700;
+  letter-spacing: .2px;
+}}
+#hero p {{
+  margin: 0;
+  opacity: .95;
+}}
+.gr-tabitem, .gr-block.gr-group, .gr-panel {{
+  background: var(--panel);
+  border-radius: 16px !important;
+  box-shadow: 0 6px 18px rgba(0,0,0,.28);
+  border: 1px solid rgba(255,255,255,.04);
+}}
+.gr-button {{
+  border-radius: 12px !important;
+  border: 1px solid rgba(255,255,255,.08) !important;
+}}
+.gradio-container .tabs .tab-nav button.selected {{
+  background: rgba(255,255,255,.06);
+  border-radius: 12px;
+  border: 1px solid rgba(255,255,255,.08);
+}}
+.badge {{
+  display:inline-block; padding:2px 8px; border-radius:999px;
+  background: rgba(255,255,255,.12); color:#fff; font-size:12px
 }}
-#brandbar strong {{ letter-spacing: .3px; }}
 """
 
-with gr.Blocks(css=THEME_CSS, title="Foley Studio · ZeroGPU") as demo:
+with gr.Blocks(css=THEME_CSS, title=APP_TITLE, analytics_enabled=False) as demo:
     with gr.Row():
-        gr.HTML(f'<div id="brandbar"><strong>{BILS_BRAND}</strong> — HunyuanVideo-Foley on ZeroGPU</div>')
+        gr.HTML(f"""
+        <div id="hero">
+          <h1>{APP_TITLE}</h1>
+          <p>{APP_TAGLINE}</p>
+          <div style="margin-top:8px"><span class="badge">ZeroGPU</span> <span class="badge">Auto-trim ≤ {MAX_SECS}s</span> <span class="badge">Downscale {TARGET_H}p</span></div>
+        </div>
+        """)
 
     with gr.Tabs():
         with gr.Tab("🎬 Single Clip"):
             with gr.Group():
-                project_name = gr.Textbox(label="Project name (optional)", placeholder="e.g., JawharaFM Teaser 09-2025")
+                project_name = gr.Textbox(
+                    label="Project name (optional)",
+                    placeholder="Enter a short label for this clip"
+                )
                 with gr.Row():
                     v_single = gr.Video(label=f"Video (≤ ~{MAX_SECS}s recommended)")
-                    p_single = gr.Textbox(label="Sound prompt (optional)", placeholder="e.g., soft footsteps, indoor reverb, light rain")
+                    p_single = gr.Textbox(
+                        label="Sound prompt (optional)",
+                        placeholder="e.g., soft footsteps on wood, light rain, indoor reverb"
+                    )
                 with gr.Row():
-                    want_mux_single = gr.Checkbox(value=True, label="Mux foley back into video (MP4)")
+                    want_mux_single = gr.Checkbox(value=True, label="Mux foley into MP4 output")
                 run_btn = gr.Button("Generate", variant="primary")
                 with gr.Row():
                     out_audio = gr.Audio(label=f"Generated Foley ({SR//1000} kHz WAV)", type="filepath")
                     out_mux = gr.Video(label="Video + Foley (MP4)", visible=True)
                 status_md = gr.Markdown()
-                history_table = gr.Dataframe(headers=["Step", "Note"], datatype=["str","str"], interactive=False, wrap=True, label="Activity")
+                history_table = gr.Dataframe(
+                    headers=["Step", "Note"], datatype=["str","str"],
+                    interactive=False, wrap=True, label="Activity", height=200
+                )
 
             run_btn.click(
                 single_generate,
@@ -294,7 +343,10 @@ with gr.Blocks(css=THEME_CSS, title="Foley Studio · ZeroGPU") as demo:
             want_mux_b = gr.Checkbox(value=True, label="Mux each output")
             go_b = gr.Button("Run batch-lite")
             batch_status = gr.Markdown()
-            batch_log = gr.Dataframe(headers=["Step","Note"], datatype=["str","str"], interactive=False, wrap=True, label="Batch Log")
+            batch_log = gr.Dataframe(
+                headers=["Step","Note"], datatype=["str","str"],
+                interactive=False, wrap=True, label="Batch Log", height=240
+            )
 
             go_b.click(
                 batch_lite_generate,
@@ -302,17 +354,16 @@ with gr.Blocks(css=THEME_CSS, title="Foley Studio · ZeroGPU") as demo:
                 outputs=[batch_status, batch_log]
             )
 
-        with gr.Tab("⚙️ Settings / Tips"):
+        with gr.Tab("ℹ️ Tips"):
             gr.Markdown(f"""
-**ZeroGPU Tips**
-- Clips are trimmed to **≤ {MAX_SECS}s** automatically.
-- Video downscaled to **{TARGET_H}p** to fit the GPU time window.
-- If you hit a quota error, try again later (ZeroGPU limits minutes per visitor).
-
-**Branding via ENV**
-- `BILS_BRAND` → header text
-- `PRIMARY_COLOR` → hex color
-- `MAX_SECS`, `TARGET_H`, `TARGET_SR` → processing behavior
+**Usage guidelines**
+- Keep clips short (the tool trims to **≤ {MAX_SECS}s** automatically).
+- The video is downscaled to **{TARGET_H}p** to fit the ZeroGPU time window.
+- If you see a quota message, just try again later (ZeroGPU limits GPU minutes per visitor).
+
+**Outputs**
+- WAV is **{SR//1000} kHz** stereo.
+- Enable **Mux** to get a ready MP4 with the generated foley track.
 """)
 
     demo.queue(max_size=24).launch()