updated images with ai images

2026-05-16 19:45:58 +02:00
parent 18193f88fc
commit 581b94eb0c
91 changed files with 960 additions and 2046 deletions
@@ -0,0 +1,125 @@
+"""Single test clip — corrected WanImageToVideo workflow."""
+import json, time, urllib.request, os, random
+
+COMFY = "http://localhost:8188"
+IMAGE_PATH = "assets/images/hero/hero-carpet-cleaning.webp"
+OUT_DIR = "assets/videos/clips"
+os.makedirs(OUT_DIR, exist_ok=True)
+
+
+def upload_image(image_path):
+    fname = os.path.basename(image_path)
+    with open(image_path, "rb") as f:
+        img_data = f.read()
+    boundary = "----FormBoundary123456"
+    body = (
+        f"--{boundary}\r\n"
+        f'Content-Disposition: form-data; name="image"; filename="{fname}"\r\n'
+        f"Content-Type: image/webp\r\n\r\n"
+    ).encode() + img_data + f"\r\n--{boundary}--\r\n".encode()
+    req = urllib.request.Request(
+        f"{COMFY}/upload/image", data=body,
+        headers={"Content-Type": f"multipart/form-data; boundary={boundary}"},
+    )
+    with urllib.request.urlopen(req) as resp:
+        result = json.loads(resp.read())
+    print(f"  uploaded: {result['name']}")
+    return result["name"]
+
+
+def build_workflow(image_name, prompt, frames=25):
+    # WanImageToVideo is a conditioning node, NOT a sampler.
+    # outputs: [0]=positive CONDITIONING, [1]=negative CONDITIONING, [2]=latent LATENT
+    # start_image is optional IMAGE — anchors first frame.
+    return {
+        "1": {"class_type": "UnetLoaderGGUF", "inputs": {"unet_name": "Wan2.2-TI2V-5B-Q4_K_M.gguf"}},
+        "2": {"class_type": "CLIPLoader", "inputs": {"clip_name": "umt5_xxl_fp8_e4m3fn_scaled.safetensors", "type": "wan"}},
+        "3": {"class_type": "VAELoader", "inputs": {"vae_name": "wan_2.1_vae.safetensors"}},
+        "4": {"class_type": "LoadImage", "inputs": {"image": image_name}},
+        "5": {"class_type": "CLIPTextEncode", "inputs": {"clip": ["2", 0], "text": prompt}},
+        "6": {"class_type": "CLIPTextEncode", "inputs": {"clip": ["2", 0], "text": "blur, low quality, distortion, text, watermark, people, jitter"}},
+        "7": {
+            "class_type": "WanImageToVideo",
+            "inputs": {
+                "positive":    ["5", 0],
+                "negative":    ["6", 0],
+                "vae":         ["3", 0],
+                "start_image": ["4", 0],
+                "width": 832, "height": 480, "length": frames, "batch_size": 1,
+            },
+        },
+        "8": {
+            "class_type": "KSampler",
+            "inputs": {
+                "model":        ["1", 0],
+                "positive":     ["7", 0],
+                "negative":     ["7", 1],
+                "latent_image": ["7", 2],
+                "seed": 42, "steps": 20, "cfg": 6.0,
+                "sampler_name": "uni_pc", "scheduler": "simple", "denoise": 1.0,
+            },
+        },
+        # VAEDecodeTiled handles video (5D) latents — VAEDecode only handles images (4D)
+        "9": {"class_type": "VAEDecodeTiled", "inputs": {"samples": ["8", 0], "vae": ["3", 0], "tile_size": 512, "overlap": 64, "temporal_size": 64, "temporal_overlap": 8}},
+        "10": {
+            "class_type": "SaveAnimatedWEBP",
+            "inputs": {"images": ["9", 0], "filename_prefix": "wan_test", "fps": 12, "lossless": False, "quality": 85, "method": "default"},
+        },
+    }
+
+
+def queue_prompt(workflow):
+    data = json.dumps({"prompt": workflow}).encode()
+    req = urllib.request.Request(f"{COMFY}/prompt", data=data, headers={"Content-Type": "application/json"})
+    with urllib.request.urlopen(req) as resp:
+        return json.loads(resp.read())["prompt_id"]
+
+
+def wait_for_result(prompt_id, timeout=1800):
+    start = time.time()
+    while time.time() - start < timeout:
+        with urllib.request.urlopen(f"{COMFY}/history/{prompt_id}") as resp:
+            hist = json.loads(resp.read())
+        if prompt_id in hist:
+            entry = hist[prompt_id]
+            if entry.get("status", {}).get("status_str") == "error":
+                print(f"  ERROR: {entry['status'].get('messages', '')}")
+                return None
+            for node_out in entry.get("outputs", {}).values():
+                if "gifs" in node_out:
+                    return node_out["gifs"]
+                if "images" in node_out:
+                    return node_out["images"]
+        elapsed = int(time.time() - start)
+        print(f"  waiting... {elapsed}s", flush=True)
+        time.sleep(15)
+    return None
+
+
+def download_output(vid_info, out_path):
+    fname = vid_info["filename"]
+    subfolder = vid_info.get("subfolder", "")
+    img_type = vid_info.get("type", "output")
+    url = f"{COMFY}/view?filename={fname}&subfolder={subfolder}&type={img_type}"
+    with urllib.request.urlopen(url) as resp:
+        data = resp.read()
+    with open(out_path, "wb") as f:
+        f.write(data)
+    print(f"  saved: {out_path} ({len(data)//1024}KB)")
+
+
+print("[TEST v2] WanImageToVideo → KSampler → VAEDecode → SaveAnimatedWEBP")
+image_name = upload_image(IMAGE_PATH)
+workflow = build_workflow(
+    image_name,
+    "slow dolly forward across clean plush cream carpet, gentle camera push toward the far wall, warm afternoon light, cinematic smooth motion",
+    frames=9,
+)
+prompt_id = queue_prompt(workflow)
+print(f"  queued: {prompt_id}")
+results = wait_for_result(prompt_id)
+if results:
+    download_output(results[0], f"{OUT_DIR}/test-clip-01.webp")
+    print("SUCCESS")
+else:
+    print("FAILED")