單圖產生汽車影片
上傳車輛照片,產生多角度高品質影片。
下載工作流程在 Cloud 中試用 API 文件
關於
這個 ComfyUI 工作流程可將單張車輛圖像轉換為多角度展示的高品質影片。透過 ByteDanceSeedreamNode 與 Kling 模型等進階節點,流程可合成模擬專業攝影棚拍攝的真實感影片。流程以 LoadImage 節點匯入車輛照片,接著由 ByteDanceSeedreamNode 處理圖像,產生模擬不同視角的多張影格,最後用 CreateVideo 節點將這些影格組合成流暢影片。
此流程最大特色是能自動化從單張圖像產生專業級汽車產品影片。SaveImage 與 SaveVideo 節點可分別儲存單張影格與最終影片。Reroute 與 BatchImagesNode 節點則讓資料處理更順暢,提升流程穩定性與易用性。此技術特別適合汽車行銷與電商平台,用動態視覺內容強化產品頁面。
您提供的內容
- 正在讀取工作流程
在 ComfyUI 中執行。 此工作流程無法在此頁面上執行。請在已載入所有內容的 Comfy Cloud 中開啟,或下載到您自己的 ComfyUI。
在 Cloud 中試用範例
工作流程內部
節點圖的唯讀預覽。請在 Comfy Cloud 中開啟,或下載後編輯節點。
Loading the graph
Open full-size workflow preview- 作者
- Rob
需要自訂節點
- 正在讀取工作流程
從程式碼使用此工作流程
可行時會根據工作流程圖和輸入內容產生;否則會顯示儲存的範例。執行可能產生費用的程式碼前,請檢查認證資訊和端點。
1 取得 API 金鑰 ↗
What you need
- Runs on
- Comfy Cloud
- 金鑰
- COMFY_API_KEY
需要付費 Cloud 方案與點數。
下載 API JSONAPI 文件 ↗
# Python 3.10+: pip install comfy-sdk==0.4.0
import os
from comfy_sdk import Comfy
api_key = os.environ.get("COMFY_API_KEY", "").strip()
if not api_key: raise RuntimeError("Set COMFY_API_KEY")
os.environ.setdefault("COMFY_BASE_URL", "https://cloud.comfy.org")
client = Comfy(api_key=api_key)
workflow = client.workflows.from_json({
"10": {
"class_type": "LoadImage",
"inputs": {
"image": "product_car.jpg"
},
"_meta": {
"title": "Load Image"
}
},
"11": {
"class_type": "ByteDanceSeedreamNode",
"inputs": {
"fail_on_partial": True,
"height": 2048,
"image": [
"10",
0
],
"max_images": 1,
"model": "seedream-4-5-251128",
"prompt": "Create an extreme close up view of the main element in the scene. Studio scene with a white background.",
"seed": 1,
"sequential_image_generation": "disabled",
"size_preset": "2560x1440 (16:9)",
"watermark": False,
"width": 2048
},
"_meta": {
"title": "ByteDance Seedream 4.5 & 5.0"
}
},
"13": {
"class_type": "SaveImage",
"inputs": {
"filename_prefix": "ComfyUI-image_2",
"images": [
"14",
0
]
},
"_meta": {
"title": "Save Image"
}
},
"14": {
"class_type": "ByteDanceSeedreamNode",
"inputs": {
"fail_on_partial": True,
"height": 2048,
"image": [
"10",
0
],
"max_images": 1,
"model": "seedream-4-5-251128",
"prompt": "Create a low view, super-wide shot of the main subject in this scene. Studio scene with a white background.",
"seed": 2,
"sequential_image_generation": "disabled",
"size_preset": "2560x1440 (16:9)",
"watermark": False,
"width": 2048
},
"_meta": {
"title": "ByteDance Seedream 4.5 & 5.0"
}
},
"16": {
"class_type": "ByteDanceSeedreamNode",
"inputs": {
"fail_on_partial": True,
"height": 2048,
"image": [
"10",
0
],
"max_images": 1,
"model": "seedream-4-5-251128",
"prompt": "Side view of the main subject in the scene. Studio scene with a white background.",
"seed": 3,
"sequential_image_generation": "disabled",
"size_preset": "2560x1440 (16:9)",
"watermark": False,
"width": 2048
},
"_meta": {
"title": "ByteDance Seedream 4.5 & 5.0"
}
},
"18": {
"class_type": "ByteDanceSeedreamNode",
"inputs": {
"fail_on_partial": True,
"height": 2048,
"image": [
"10",
0
],
"max_images": 1,
"model": "seedream-4-5-251128",
"prompt": "Create an extreme wide shot from behind the subject near the passenger side of the back rims. Studio scene with a white background.",
"seed": 4,
"sequential_image_generation": "disabled",
"size_preset": "2560x1440 (16:9)",
"watermark": False,
"width": 2048
},
"_meta": {
"title": "ByteDance Seedream 4.5 & 5.0"
}
},
"22": {
"class_type": "SaveVideo",
"inputs": {
"filename_prefix": "video/ComfyUI-video_1",
"format": "auto",
"format.codec": "auto",
"video": [
"42:42",
0
]
},
"_meta": {
"title": "Save Video"
}
},
"23": {
"class_type": "SaveImage",
"inputs": {
"filename_prefix": "ComfyUI-image_1",
"images": [
"11",
0
]
},
"_meta": {
"title": "Save Image"
}
},
"24": {
"class_type": "SaveImage",
"inputs": {
"filename_prefix": "ComfyUI-image_3",
"images": [
"16",
0
]
},
"_meta": {
"title": "Save Image"
}
},
"25": {
"class_type": "SaveImage",
"inputs": {
"filename_prefix": "ComfyUI-image_4",
"images": [
"18",
0
]
},
"_meta": {
"title": "Save Image"
}
},
"26": {
"class_type": "SaveVideo",
"inputs": {
"filename_prefix": "video/ComfyUI-video_2",
"format": "auto",
"format.codec": "auto",
"video": [
"58:42",
0
]
},
"_meta": {
"title": "Save Video"
}
},
"27": {
"class_type": "SaveVideo",
"inputs": {
"filename_prefix": "video/ComfyUI-video_3",
"format": "auto",
"format.codec": "auto",
"video": [
"59:42",
0
]
},
"_meta": {
"title": "Save Video"
}
},
"47": {
"class_type": "CreateVideo",
"inputs": {
"bit_depth": "auto",
"codec": "none",
"color_space": "sRGB",
"fps": [
"51",
2
],
"images": [
"54",
0
]
},
"_meta": {
"title": "Create Video"
}
},
"48": {
"class_type": "SaveVideo",
"inputs": {
"filename_prefix": "video/ComfyUI",
"format": "auto",
"format.codec": "auto",
"video": [
"47",
0
]
},
"_meta": {
"title": "Save Video"
}
},
"49": {
"class_type": "GetVideoComponents",
"inputs": {
"video": [
"58:42",
0
]
},
"_meta": {
"title": "Get Video Components"
}
},
"50": {
"class_type": "GetVideoComponents",
"inputs": {
"video": [
"59:42",
0
]
},
"_meta": {
"title": "Get Video Components"
}
},
"51": {
"class_type": "GetVideoComponents",
"inputs": {
"video": [
"42:42",
0
]
},
"_meta": {
"title": "Get Video Components"
}
},
"54": {
"class_type": "BatchImagesNode",
"inputs": {
"images.image0": [
"51",
0
],
"images.image1": [
"49",
0
],
"images.image2": [
"50",
0
]
},
"_meta": {
"title": "Batch Images"
}
},
"42:42": {
"class_type": "KlingStartEndFrameNode",
"inputs": {
"aspect_ratio": "16:9",
"cfg_scale": 0.5,
"end_frame": [
"14",
0
],
"mode": "pro mode / 5s duration / kling-v2-5-turbo",
"negative_prompt": [
"42:48",
0
],
"prompt": [
"42:49",
0
],
"start_frame": [
"11",
0
]
},
"_meta": {
"title": "Kling Start-End Frame to Video"
}
},
"42:45": {
"class_type": "BatchImagesNode",
"inputs": {
"images.image0": [
"11",
0
],
"images.image1": [
"14",
0
]
},
"_meta": {
"title": "Batch Images"
}
},
"42:46": {
"class_type": "PreviewAny",
"inputs": {
"source": [
"42:49",
0
]
},
"_meta": {
"title": "Preview Positive Prompt"
}
},
"42:47": {
"class_type": "PreviewAny",
"inputs": {
"source": [
"42:48",
0
]
},
"_meta": {
"title": "Preview Negative Prompt"
}
},
"42:48": {
"class_type": "RegexExtract",
"inputs": {
"case_insensitive": True,
"dotall": False,
"group_index": 1,
"mode": "First Match",
"multiline": False,
"regex_pattern": "(?<=--no\\s)(.+)$",
"string": [
"42:50",
0
]
},
"_meta": {
"title": "Extract Text"
}
},
"42:49": {
"class_type": "RegexExtract",
"inputs": {
"case_insensitive": True,
"dotall": False,
"group_index": 1,
"mode": "First Match",
"multiline": False,
"regex_pattern": "^(.+?)(?=\\s+--no)",
"string": [
"42:50",
0
]
},
"_meta": {
"title": "Extract Text"
}
},
"42:50": {
"class_type": "GeminiNode",
"inputs": {
"images": [
"42:45",
0
],
"model": "gemini-2.5-flash",
"prompt": "STEP 1: ANALYZE BOTH IMAGES\nBefore crafting the prompt, analyze both images to understand:\n\n\nCamera distance (wide shot, medium, close-up, extreme close-up, macro)\nCamera angle (low-angle, high-angle, eye-level, overhead, Dutch, worm's eye, bird's eye)\nCamera position relative to subject (front, side profile, 3/4 view, back, overhead)\nFraming (full body, torso, chest-up, face only, detail/macro)\nSubject position and pose\nLighting setup and direction\nBackground elements and depth\nFocal length characteristics (wide-angle distortion vs telephoto compression)\n\nSTEP 2: DETERMINE CAMERA MOVEMENT\nBased on the differences between images, identify the primary camera movement:\n\n\nDistance change: Dolly-in (closer) or Dolly-out (farther)\nVertical angle change: Tilt up/down or Crane up/down\nHorizontal angle change: Pan left/right or Orbit/Arc around subject\nHeight change: Crane/Pedestal up or down\nLateral movement: Truck/Track left or right\n\nSTEP 3: OUTPUT ONLY THE PROMPT\nDo not include explanations, commentary, or additional text.\nOutput ONLY the transition prompt in this format:\n\n\n[CAMERA MOVEMENT] from [START POSITION] to [END POSITION]. Subject remains stationary and consistent throughout. Camera [MOVEMENT VERB] fluidly, maintaining [FOCUS BEHAVIOR] on [KEY ELEMENT]. Lighting stays constant, background [BACKGROUND BEHAVIOR]. [TECHNICAL SPECS]. --no [NEGATIVE ELEMENTS]\n\nCore Principle\n\nFor transitions where the subject/scene remains consistent and only the camera position changes, use this structured approach to ensure smooth, cinematic results.\n\n\nPrompt Structure Template\n\n[CAMERA MOVEMENT] from [START POSITION] to [END POSITION]. Subject remains stationary and consistent throughout. Camera [MOVEMENT VERB] fluidly, maintaining [FOCUS BEHAVIOR] on [KEY ELEMENT]. Lighting stays constant, background [BACKGROUND BEHAVIOR]. [TECHNICAL SPECS]. --no [NEGATIVE ELEMENTS]\n\nComponent Breakdown\n\n1. CAMERA MOVEMENT TYPES\n\n\nDolly-in/Push-in: Camera moves forward toward subject\nDolly-out/Pull-back: Camera moves backward away from subject\nOrbit/Arc: Camera circles around subject (specify left/right)\nTilt: Camera angles up or down on fixed position\nPan: Camera swivels left or right horizontally\nCrane/Pedestal: Camera moves vertically up or down\nTracking/Trucking: Camera moves parallel to subject\n\n2. POSITION DESCRIPTORS\n\n\nDistance: wide shot, medium shot, close-up, extreme close-up, macro\nAngle: low-angle, high-angle, eye-level, overhead, Dutch angle, worm's eye view, bird's eye view\nFraming: full body, three-quarter, torso, chest-up, face, detail\n\n3. MOVEMENT VERBS (choose intensity)\n\n\nSlowly, steadily, smoothly, gradually, fluidly\nRapidly, dynamically, swiftly\nElegantly, dramatically, confidently\n\n4. FOCUS BEHAVIOR\n\n\nSharp focus throughout\nRack focus from [A] to [B]\nMaintains focus on [element]\nShallow depth of field\nDeep focus\nProgressive focus shift\n\n5. BACKGROUND BEHAVIOR\n\n\nStays consistent\nBlurs to bokeh\nSharpens into view\nFades to [color]\nRemains sharp throughout\n\n6. TECHNICAL SPECS (standard cinematic quality)\n\n\nCinematic depth of field\nStable gimbal/dolly movement\nProfessional commercial grade\n24fps (or 30fps, 60fps for slow-motion feel)\nClean sensor, no grain (or \"film grain\" if desired)\n\n\nExample Prompts by Movement Type\n\nDOLLY-IN (Wide → Close-up)\n\n\nSmooth camera dolly-in from wide establishing shot to extreme close-up. Subject remains stationary. Camera pushes forward steadily, transitioning focus from [background element] to [detail]. Lighting stays constant, background softly blurs to bokeh. Cinematic depth of field, stable gimbal movement, 24fps. --no shake, cuts, warping, distortion\n\nDOLLY-OUT (Close-up → Wide)\n\n\nSmooth camera pull-back from extreme close-up to wide establishing shot. Subject remains centered and stationary. Camera retreats steadily, revealing environment progressively. Focus maintains on subject while background sharpens into view. Even lighting, cinematic depth of field, 24fps. --no shake, cuts, jump, warping\n\nORBIT (Circular movement)\n\n\nSmooth camera orbit from front view to side profile. Subject remains stationary at center. Camera arcs left fluidly in 90-degree rotation, maintaining constant distance and eye-level angle. Lighting stays consistent, sharp focus throughout. Gimbal stabilized circular tracking, 24fps. --no shake, warping, speed changes, distortion\n\nTILT (Vertical angle change)\n\n\nSmooth camera tilt from low-angle to high-angle. Subject remains stationary. Camera tilts upward gradually, shifting perspective from worm's eye view to bird's eye view. Distance stays constant, lighting consistent. Stable tripod tilt, cinematic framing, 24fps. --no shake, panning, horizontal movement, warping\n\nPAN (Horizontal swivel)\n\n\nSmooth camera pan from left profile to right profile. Subject remains stationary. Camera swivels horizontally across 180 degrees, maintaining constant distance and eye-level. Sharp focus throughout, even lighting. Fluid tripod pan, 24fps. --no shake, tilting, vertical movement, cuts\n\nCRANE UP/DOWN\n\n\nSmooth camera crane from ground-level to overhead view. Subject remains stationary at center. Camera elevates vertically, angle shifting from eye-level to bird's eye view. Distance adjusts proportionally, lighting consistent. Stabilized crane movement, 24fps. --no shake, lateral movement, warping, distortion\n\nKey Rules for Camera-Only Transitions\n\nAlways specify subject remains stationary - This prevents unwanted morphing\nDescribe start AND end positions clearly - Gives model clear targets\nUse specific movement verbs - \"Dolly-in\" is clearer than \"moves closer\"\nLock down what DOESN'T change - Lighting, subject, clothing, expression\nInclude technical quality markers - Gimbal, 24fps, cinematic, stable\nNegative prompts are critical - Prevents warping/morphing/distortion\nKeep prompts under 75 words - Model weights early words heavier\nOne primary movement per transition - Don't combine dolly + pan + tilt\n\nAdvanced Techniques\n\nSpeed Ramping:\n\"Camera begins slow dolly-in, accelerates midway, then decelerates smoothly into final close-up\"\n\n\nCombined Movements (use sparingly):\n\"Camera dollies in while simultaneously tilting down, maintaining subject center-frame\"\n\n\nFocus Pulls:\n\"Rack focus from foreground [element] to background [element] as camera remains locked\"\n\n\nLens Simulation:\n\"Wide-angle 24mm distortion gradually compresses to telephoto 85mm compression\"\n\n\n\nCritical Output Rule\n\nOUTPUT ONLY THE PROMPT. NO EXPLANATIONS. NO COMMENTARY. NO ADDITIONAL TEXT.",
"seed": 123,
"system_prompt": ""
},
"_meta": {
"title": "Google Gemini"
}
},
"58:42": {
"class_type": "KlingStartEndFrameNode",
"inputs": {
"aspect_ratio": "16:9",
"cfg_scale": 0.5,
"end_frame": [
"16",
0
],
"mode": "pro mode / 5s duration / kling-v2-5-turbo",
"negative_prompt": [
"58:48",
0
],
"prompt": [
"58:49",
0
],
"start_frame": [
"14",
0
]
},
"_meta": {
"title": "Kling Start-End Frame to Video"
}
},
"58:45": {
"class_type": "BatchImagesNode",
"inputs": {
"images.image0": [
"14",
0
],
"images.image1": [
"16",
0
]
},
"_meta": {
"title": "Batch Images"
}
},
"58:46": {
"class_type": "PreviewAny",
"inputs": {
"source": [
"58:49",
0
]
},
"_meta": {
"title": "Preview Positive Prompt"
}
},
"58:47": {
"class_type": "PreviewAny",
"inputs": {
"source": [
"58:48",
0
]
},
"_meta": {
"title": "Preview Negative Prompt"
}
},
"58:48": {
"class_type": "RegexExtract",
"inputs": {
"case_insensitive": True,
"dotall": False,
"group_index": 1,
"mode": "First Match",
"multiline": False,
"regex_pattern": "(?<=--no\\s)(.+)$",
"string": [
"58:50",
0
]
},
"_meta": {
"title": "Extract Text"
}
},
"58:49": {
"class_type": "RegexExtract",
"inputs": {
"case_insensitive": True,
"dotall": False,
"group_index": 1,
"mode": "First Match",
"multiline": False,
"regex_pattern": "^(.+?)(?=\\s+--no)",
"string": [
"58:50",
0
]
},
"_meta": {
"title": "Extract Text"
}
},
"58:50": {
"class_type": "GeminiNode",
"inputs": {
"images": [
"58:45",
0
],
"model": "gemini-2.5-flash",
"prompt": "STEP 1: ANALYZE BOTH IMAGES\nBefore crafting the prompt, analyze both images to understand:\n\n\nCamera distance (wide shot, medium, close-up, extreme close-up, macro)\nCamera angle (low-angle, high-angle, eye-level, overhead, Dutch, worm's eye, bird's eye)\nCamera position relative to subject (front, side profile, 3/4 view, back, overhead)\nFraming (full body, torso, chest-up, face only, detail/macro)\nSubject position and pose\nLighting setup and direction\nBackground elements and depth\nFocal length characteristics (wide-angle distortion vs telephoto compression)\n\nSTEP 2: DETERMINE CAMERA MOVEMENT\nBased on the differences between images, identify the primary camera movement:\n\n\nDistance change: Dolly-in (closer) or Dolly-out (farther)\nVertical angle change: Tilt up/down or Crane up/down\nHorizontal angle change: Pan left/right or Orbit/Arc around subject\nHeight change: Crane/Pedestal up or down\nLateral movement: Truck/Track left or right\n\nSTEP 3: OUTPUT ONLY THE PROMPT\nDo not include explanations, commentary, or additional text.\nOutput ONLY the transition prompt in this format:\n\n\n[CAMERA MOVEMENT] from [START POSITION] to [END POSITION]. Subject remains stationary and consistent throughout. Camera [MOVEMENT VERB] fluidly, maintaining [FOCUS BEHAVIOR] on [KEY ELEMENT]. Lighting stays constant, background [BACKGROUND BEHAVIOR]. [TECHNICAL SPECS]. --no [NEGATIVE ELEMENTS]\n\nCore Principle\n\nFor transitions where the subject/scene remains consistent and only the camera position changes, use this structured approach to ensure smooth, cinematic results.\n\n\nPrompt Structure Template\n\n[CAMERA MOVEMENT] from [START POSITION] to [END POSITION]. Subject remains stationary and consistent throughout. Camera [MOVEMENT VERB] fluidly, maintaining [FOCUS BEHAVIOR] on [KEY ELEMENT]. Lighting stays constant, background [BACKGROUND BEHAVIOR]. [TECHNICAL SPECS]. --no [NEGATIVE ELEMENTS]\n\nComponent Breakdown\n\n1. CAMERA MOVEMENT TYPES\n\n\nDolly-in/Push-in: Camera moves forward toward subject\nDolly-out/Pull-back: Camera moves backward away from subject\nOrbit/Arc: Camera circles around subject (specify left/right)\nTilt: Camera angles up or down on fixed position\nPan: Camera swivels left or right horizontally\nCrane/Pedestal: Camera moves vertically up or down\nTracking/Trucking: Camera moves parallel to subject\n\n2. POSITION DESCRIPTORS\n\n\nDistance: wide shot, medium shot, close-up, extreme close-up, macro\nAngle: low-angle, high-angle, eye-level, overhead, Dutch angle, worm's eye view, bird's eye view\nFraming: full body, three-quarter, torso, chest-up, face, detail\n\n3. MOVEMENT VERBS (choose intensity)\n\n\nSlowly, steadily, smoothly, gradually, fluidly\nRapidly, dynamically, swiftly\nElegantly, dramatically, confidently\n\n4. FOCUS BEHAVIOR\n\n\nSharp focus throughout\nRack focus from [A] to [B]\nMaintains focus on [element]\nShallow depth of field\nDeep focus\nProgressive focus shift\n\n5. BACKGROUND BEHAVIOR\n\n\nStays consistent\nBlurs to bokeh\nSharpens into view\nFades to [color]\nRemains sharp throughout\n\n6. TECHNICAL SPECS (standard cinematic quality)\n\n\nCinematic depth of field\nStable gimbal/dolly movement\nProfessional commercial grade\n24fps (or 30fps, 60fps for slow-motion feel)\nClean sensor, no grain (or \"film grain\" if desired)\n\n\nExample Prompts by Movement Type\n\nDOLLY-IN (Wide → Close-up)\n\n\nSmooth camera dolly-in from wide establishing shot to extreme close-up. Subject remains stationary. Camera pushes forward steadily, transitioning focus from [background element] to [detail]. Lighting stays constant, background softly blurs to bokeh. Cinematic depth of field, stable gimbal movement, 24fps. --no shake, cuts, warping, distortion\n\nDOLLY-OUT (Close-up → Wide)\n\n\nSmooth camera pull-back from extreme close-up to wide establishing shot. Subject remains centered and stationary. Camera retreats steadily, revealing environment progressively. Focus maintains on subject while background sharpens into view. Even lighting, cinematic depth of field, 24fps. --no shake, cuts, jump, warping\n\nORBIT (Circular movement)\n\n\nSmooth camera orbit from front view to side profile. Subject remains stationary at center. Camera arcs left fluidly in 90-degree rotation, maintaining constant distance and eye-level angle. Lighting stays consistent, sharp focus throughout. Gimbal stabilized circular tracking, 24fps. --no shake, warping, speed changes, distortion\n\nTILT (Vertical angle change)\n\n\nSmooth camera tilt from low-angle to high-angle. Subject remains stationary. Camera tilts upward gradually, shifting perspective from worm's eye view to bird's eye view. Distance stays constant, lighting consistent. Stable tripod tilt, cinematic framing, 24fps. --no shake, panning, horizontal movement, warping\n\nPAN (Horizontal swivel)\n\n\nSmooth camera pan from left profile to right profile. Subject remains stationary. Camera swivels horizontally across 180 degrees, maintaining constant distance and eye-level. Sharp focus throughout, even lighting. Fluid tripod pan, 24fps. --no shake, tilting, vertical movement, cuts\n\nCRANE UP/DOWN\n\n\nSmooth camera crane from ground-level to overhead view. Subject remains stationary at center. Camera elevates vertically, angle shifting from eye-level to bird's eye view. Distance adjusts proportionally, lighting consistent. Stabilized crane movement, 24fps. --no shake, lateral movement, warping, distortion\n\nKey Rules for Camera-Only Transitions\n\nAlways specify subject remains stationary - This prevents unwanted morphing\nDescribe start AND end positions clearly - Gives model clear targets\nUse specific movement verbs - \"Dolly-in\" is clearer than \"moves closer\"\nLock down what DOESN'T change - Lighting, subject, clothing, expression\nInclude technical quality markers - Gimbal, 24fps, cinematic, stable\nNegative prompts are critical - Prevents warping/morphing/distortion\nKeep prompts under 75 words - Model weights early words heavier\nOne primary movement per transition - Don't combine dolly + pan + tilt\n\nAdvanced Techniques\n\nSpeed Ramping:\n\"Camera begins slow dolly-in, accelerates midway, then decelerates smoothly into final close-up\"\n\n\nCombined Movements (use sparingly):\n\"Camera dollies in while simultaneously tilting down, maintaining subject center-frame\"\n\n\nFocus Pulls:\n\"Rack focus from foreground [element] to background [element] as camera remains locked\"\n\n\nLens Simulation:\n\"Wide-angle 24mm distortion gradually compresses to telephoto 85mm compression\"\n\n\n\nCritical Output Rule\n\nOUTPUT ONLY THE PROMPT. NO EXPLANATIONS. NO COMMENTARY. NO ADDITIONAL TEXT.",
"seed": 123,
"system_prompt": ""
},
"_meta": {
"title": "Google Gemini"
}
},
"59:42": {
"class_type": "KlingStartEndFrameNode",
"inputs": {
"aspect_ratio": "16:9",
"cfg_scale": 0.5,
"end_frame": [
"18",
0
],
"mode": "pro mode / 5s duration / kling-v2-5-turbo",
"negative_prompt": [
"59:48",
0
],
"prompt": [
"59:49",
0
],
"start_frame": [
"16",
0
]
},
"_meta": {
"title": "Kling Start-End Frame to Video"
}
},
"59:45": {
"class_type": "BatchImagesNode",
"inputs": {
"images.image0": [
"16",
0
],
"images.image1": [
"18",
0
]
},
"_meta": {
"title": "Batch Images"
}
},
"59:46": {
"class_type": "PreviewAny",
"inputs": {
"source": [
"59:49",
0
]
},
"_meta": {
"title": "Preview Positive Prompt"
}
},
"59:47": {
"class_type": "PreviewAny",
"inputs": {
"source": [
"59:48",
0
]
},
"_meta": {
"title": "Preview Negative Prompt"
}
},
"59:48": {
"class_type": "RegexExtract",
"inputs": {
"case_insensitive": True,
"dotall": False,
"group_index": 1,
"mode": "First Match",
"multiline": False,
"regex_pattern": "(?<=--no\\s)(.+)$",
"string": [
"59:50",
0
]
},
"_meta": {
"title": "Extract Text"
}
},
"59:49": {
"class_type": "RegexExtract",
"inputs": {
"case_insensitive": True,
"dotall": False,
"group_index": 1,
"mode": "First Match",
"multiline": False,
"regex_pattern": "^(.+?)(?=\\s+--no)",
"string": [
"59:50",
0
]
},
"_meta": {
"title": "Extract Text"
}
},
"59:50": {
"class_type": "GeminiNode",
"inputs": {
"images": [
"59:45",
0
],
"model": "gemini-2.5-flash",
"prompt": "STEP 1: ANALYZE BOTH IMAGES\nBefore crafting the prompt, analyze both images to understand:\n\n\nCamera distance (wide shot, medium, close-up, extreme close-up, macro)\nCamera angle (low-angle, high-angle, eye-level, overhead, Dutch, worm's eye, bird's eye)\nCamera position relative to subject (front, side profile, 3/4 view, back, overhead)\nFraming (full body, torso, chest-up, face only, detail/macro)\nSubject position and pose\nLighting setup and direction\nBackground elements and depth\nFocal length characteristics (wide-angle distortion vs telephoto compression)\n\nSTEP 2: DETERMINE CAMERA MOVEMENT\nBased on the differences between images, identify the primary camera movement:\n\n\nDistance change: Dolly-in (closer) or Dolly-out (farther)\nVertical angle change: Tilt up/down or Crane up/down\nHorizontal angle change: Pan left/right or Orbit/Arc around subject\nHeight change: Crane/Pedestal up or down\nLateral movement: Truck/Track left or right\n\nSTEP 3: OUTPUT ONLY THE PROMPT\nDo not include explanations, commentary, or additional text.\nOutput ONLY the transition prompt in this format:\n\n\n[CAMERA MOVEMENT] from [START POSITION] to [END POSITION]. Subject remains stationary and consistent throughout. Camera [MOVEMENT VERB] fluidly, maintaining [FOCUS BEHAVIOR] on [KEY ELEMENT]. Lighting stays constant, background [BACKGROUND BEHAVIOR]. [TECHNICAL SPECS]. --no [NEGATIVE ELEMENTS]\n\nCore Principle\n\nFor transitions where the subject/scene remains consistent and only the camera position changes, use this structured approach to ensure smooth, cinematic results.\n\n\nPrompt Structure Template\n\n[CAMERA MOVEMENT] from [START POSITION] to [END POSITION]. Subject remains stationary and consistent throughout. Camera [MOVEMENT VERB] fluidly, maintaining [FOCUS BEHAVIOR] on [KEY ELEMENT]. Lighting stays constant, background [BACKGROUND BEHAVIOR]. [TECHNICAL SPECS]. --no [NEGATIVE ELEMENTS]\n\nComponent Breakdown\n\n1. CAMERA MOVEMENT TYPES\n\n\nDolly-in/Push-in: Camera moves forward toward subject\nDolly-out/Pull-back: Camera moves backward away from subject\nOrbit/Arc: Camera circles around subject (specify left/right)\nTilt: Camera angles up or down on fixed position\nPan: Camera swivels left or right horizontally\nCrane/Pedestal: Camera moves vertically up or down\nTracking/Trucking: Camera moves parallel to subject\n\n2. POSITION DESCRIPTORS\n\n\nDistance: wide shot, medium shot, close-up, extreme close-up, macro\nAngle: low-angle, high-angle, eye-level, overhead, Dutch angle, worm's eye view, bird's eye view\nFraming: full body, three-quarter, torso, chest-up, face, detail\n\n3. MOVEMENT VERBS (choose intensity)\n\n\nSlowly, steadily, smoothly, gradually, fluidly\nRapidly, dynamically, swiftly\nElegantly, dramatically, confidently\n\n4. FOCUS BEHAVIOR\n\n\nSharp focus throughout\nRack focus from [A] to [B]\nMaintains focus on [element]\nShallow depth of field\nDeep focus\nProgressive focus shift\n\n5. BACKGROUND BEHAVIOR\n\n\nStays consistent\nBlurs to bokeh\nSharpens into view\nFades to [color]\nRemains sharp throughout\n\n6. TECHNICAL SPECS (standard cinematic quality)\n\n\nCinematic depth of field\nStable gimbal/dolly movement\nProfessional commercial grade\n24fps (or 30fps, 60fps for slow-motion feel)\nClean sensor, no grain (or \"film grain\" if desired)\n\n\nExample Prompts by Movement Type\n\nDOLLY-IN (Wide → Close-up)\n\n\nSmooth camera dolly-in from wide establishing shot to extreme close-up. Subject remains stationary. Camera pushes forward steadily, transitioning focus from [background element] to [detail]. Lighting stays constant, background softly blurs to bokeh. Cinematic depth of field, stable gimbal movement, 24fps. --no shake, cuts, warping, distortion\n\nDOLLY-OUT (Close-up → Wide)\n\n\nSmooth camera pull-back from extreme close-up to wide establishing shot. Subject remains centered and stationary. Camera retreats steadily, revealing environment progressively. Focus maintains on subject while background sharpens into view. Even lighting, cinematic depth of field, 24fps. --no shake, cuts, jump, warping\n\nORBIT (Circular movement)\n\n\nSmooth camera orbit from front view to side profile. Subject remains stationary at center. Camera arcs left fluidly in 90-degree rotation, maintaining constant distance and eye-level angle. Lighting stays consistent, sharp focus throughout. Gimbal stabilized circular tracking, 24fps. --no shake, warping, speed changes, distortion\n\nTILT (Vertical angle change)\n\n\nSmooth camera tilt from low-angle to high-angle. Subject remains stationary. Camera tilts upward gradually, shifting perspective from worm's eye view to bird's eye view. Distance stays constant, lighting consistent. Stable tripod tilt, cinematic framing, 24fps. --no shake, panning, horizontal movement, warping\n\nPAN (Horizontal swivel)\n\n\nSmooth camera pan from left profile to right profile. Subject remains stationary. Camera swivels horizontally across 180 degrees, maintaining constant distance and eye-level. Sharp focus throughout, even lighting. Fluid tripod pan, 24fps. --no shake, tilting, vertical movement, cuts\n\nCRANE UP/DOWN\n\n\nSmooth camera crane from ground-level to overhead view. Subject remains stationary at center. Camera elevates vertically, angle shifting from eye-level to bird's eye view. Distance adjusts proportionally, lighting consistent. Stabilized crane movement, 24fps. --no shake, lateral movement, warping, distortion\n\nKey Rules for Camera-Only Transitions\n\nAlways specify subject remains stationary - This prevents unwanted morphing\nDescribe start AND end positions clearly - Gives model clear targets\nUse specific movement verbs - \"Dolly-in\" is clearer than \"moves closer\"\nLock down what DOESN'T change - Lighting, subject, clothing, expression\nInclude technical quality markers - Gimbal, 24fps, cinematic, stable\nNegative prompts are critical - Prevents warping/morphing/distortion\nKeep prompts under 75 words - Model weights early words heavier\nOne primary movement per transition - Don't combine dolly + pan + tilt\n\nAdvanced Techniques\n\nSpeed Ramping:\n\"Camera begins slow dolly-in, accelerates midway, then decelerates smoothly into final close-up\"\n\n\nCombined Movements (use sparingly):\n\"Camera dollies in while simultaneously tilting down, maintaining subject center-frame\"\n\n\nFocus Pulls:\n\"Rack focus from foreground [element] to background [element] as camera remains locked\"\n\n\nLens Simulation:\n\"Wide-angle 24mm distortion gradually compresses to telephoto 85mm compression\"\n\n\n\nCritical Output Rule\n\nOUTPUT ONLY THE PROMPT. NO EXPLANATIONS. NO COMMENTARY. NO ADDITIONAL TEXT.",
"seed": 123,
"system_prompt": ""
},
"_meta": {
"title": "Google Gemini"
}
}
})
# run() also passes api_key to the partner nodes in this workflow.
job = client.run(workflow, api_key=api_key)問答
建議使用高解析度的 JPEG 或 PNG 格式圖像,以利最佳處理效果。
雖然流程針對汽車最佳化,但調整處理節點後也可應用於其他車輛類型。
時間取決於圖像解析度與系統效能,通常數分鐘內可完成處理與組合。
流程可產生多角度影格,但數量受限於系統運算能力與記憶體。
相關工作流程
- Seedream
Put your product in a new scene
- Seedream, MiniMax
Turn a product photo into a video
- Flux.2 Dev, Flux.2 Klein, GPT-Image-1.5, Grok, Nano Banana Pro, Qwen-Image-Edit, Seedream
1個輸入與多個編輯模型對比
- ByteDance, Seedream, Google Gemini, Nano Banana Pro
動感3D標誌動畫
- Seedream, MiniMax
產品圖片轉影片:鞋
- ByteDance, Seedream
產品場景合成與重新光照
- ByteDance, Google Gemini, Seedream
用你的產品產生海報/廣告素材