電影感橋接影片
可用現有或生成的終格,將起始影格平滑過渡到結尾,產生電影感轉場影片。
關於
電影感橋接影片是一種首尾影格轉換流程,能將單一起始圖像轉為短片,平滑過渡到指定終格。你可自備終格,也可讓流程自動生成。流程從 LoadImage 載入起始影格開始。若無終格,可在 Prompt 群組用 GeminiNode 擴展描述,再用 GeminiImage2Node 生成同一故事世界、解析度的戲劇性終格。兩張圖可用 PreviewAny 預覽,SaveImage 可儲存生成的終格以便重用。
實際動作由 Seedance 2.0 的 ByteDance2FirstLastFrameNode 負責,會在首尾影格間插值,產生平滑、電影感的過渡。你可調整格數與 FPS 控制片長與節奏,也可微調動作強度與一致性,平衡創意與主體穩定。最後 SaveVideo 寫出橋接影片,SaveImage 可選擇性儲存關鍵影格或終格。MarkdownNote 節點於畫布記錄建議設定與使用說明。
此流程特別適合需要故事連貫轉場或場景揭露時,不必從零動畫整段序列。確保兩張圖主題、光線、比例一致,有助 Seedance 2.0 產生穩定、可信的中間過渡,減少閃爍。
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
- ComfyUI
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。
What you need
- Runs on
- Comfy Cloud
- 金鑰
- COMFY_API_KEY
需要付費 Cloud 方案與點數。
# 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({
"1": {
"class_type": "LoadImage",
"inputs": {
"image": "8e8d4ef04692bc7f92093dbc8fec140ed4b89148036d863e98f8e83bdba3970b.png"
},
"_meta": {
"title": "Load Image"
}
},
"2": {
"class_type": "GeminiImage2Node",
"inputs": {
"aspect_ratio": "21:9",
"images": [
"1",
0
],
"model": "Nano Banana 2 (Gemini 3.1 Flash Image)",
"prompt": [
"8",
0
],
"resolution": "4K",
"response_modalities": "IMAGE",
"seed": 265324705690959,
"system_prompt": "You are a cinematic transition director and continuity artist for first-last-frame video generation.\n\nYour job is to create a final frame that continues the uploaded first frame into a later cinematic beat while intentionally giving the video model a big camera move to interpolate. Preserve subject identity, environment identity, lighting logic, aspect ratio, and recognizable spatial anchors. Do not preserve the original composition too tightly: by default choose a noticeably different camera position, lens scale, and viewing angle, such as a push-in, fly-through, orbit, low/high-angle reveal, over-the-shoulder pass, or close foreground parallax move.\n\nPrioritize dramatic motion potential over static continuity when they conflict, but keep the result physically plausible and connected to the same scene. The output must be a clean final frame only. No annotations, no text overlays, no storyboard marks, no logos, no UI, no captions."
},
"_meta": {
"title": "Nano Banana Pro (Google Gemini Image)"
}
},
"3": {
"class_type": "ByteDance2FirstLastFrameNode",
"inputs": {
"first_frame": [
"1",
0
],
"first_frame_asset_id": "",
"last_frame": [
"2",
0
],
"last_frame_asset_id": "",
"model": "Seedance 2.0",
"model.duration": 7,
"model.generate_audio": True,
"model.prompt": "Create a dramatic first-to-last-frame cinematic transition from the uploaded first frame to the generated final frame. Preserve subject identity, scene continuity, lighting direction, and physical realism, but make the camera travel assertively through the space instead of holding the original composition. Use a feature-film move such as a strong push-in/zoom-in, fly-through, orbit reveal, low/high-angle sweep, crash zoom, or dolly past foreground geometry. Emphasize substantial parallax, depth change, scale change, and viewpoint shift so the shot feels physically motivated, not like a crossfade or soft morph. Keep geometry stable, motion blur realistic, and action progression smooth. Do not show text, annotations, logos, UI, artifacts, or sudden unrelated scene changes.",
"model.ratio": "21:9",
"model.resolution": "480p",
"seed": 443108043,
"watermark": False
},
"_meta": {
"title": "ByteDance Seedance 2.5 First-Last-Frame to Video"
}
},
"4": {
"class_type": "SaveVideo",
"inputs": {
"filename_prefix": "video/cinematic_bridge",
"format": "mp4",
"format.codec": "h264",
"video": [
"3",
0
]
},
"_meta": {
"title": "Save Video"
}
},
"5": {
"class_type": "SaveImage",
"inputs": {
"filename_prefix": "cinematic_bridge_final_frame",
"images": [
"2",
0
]
},
"_meta": {
"title": "Save Image"
}
},
"7": {
"class_type": "GeminiNode",
"inputs": {
"images": [
"1",
0
],
"model": "gemini-3-1-flash-lite",
"prompt": "Using the provided image as the first frame, generate a clean cinematic final frame for a first-last-frame video workflow.\n\nCreate the moment 5-7 seconds later in the same story world, but make the endpoint visually bolder than a simple continuation. Keep subject identity, recognizable environment anchors, lighting direction, aspect ratio, and physical realism, while deliberately changing camera placement and framing. Default to a dramatic feature-film move: a hard push-in or zoom-in, low or high angle shift, orbit reveal, fly-through past foreground elements, crash zoom, dolly through a doorway/window/vehicle gap, or a close pass that creates strong parallax.\n\nThe last frame should be a substantially different angle or scale from the uploaded image, with clearer depth, larger foreground/background separation, and an obvious destination for the video model to travel toward. Advance the action clearly: vehicles surge forward, storms intensify, doors open, lights sweep, characters turn or move through space, smoke/fire/water/debris streams across the lens, or the camera lands on a new dramatic reveal.\n\nPreserve continuity of the world; do not swap to an unrelated location, change the main subject identity, add unrelated characters, or break plausible geometry. Keep it photorealistic and production-grade. Do not add captions, labels, arrows, annotations, logos, watermarks, random text, or UI.",
"seed": 875942294806287,
"system_prompt": "You are a cinematic transition director and continuity artist for first-last-frame video generation.\n\nYour job is to create a final frame that continues the uploaded first frame into a later cinematic beat while intentionally giving the video model a big camera move to interpolate. Preserve subject identity, environment identity, lighting logic, aspect ratio, and recognizable spatial anchors. Do not preserve the original composition too tightly: by default choose a noticeably different camera position, lens scale, and viewing angle, such as a push-in, fly-through, orbit, low/high-angle reveal, over-the-shoulder pass, or close foreground parallax move.\n\nPrioritize dramatic motion potential over static continuity when they conflict, but keep the result physically plausible and connected to the same scene. The output must be a clean final frame only. No annotations, no text overlays, no storyboard marks, no logos, no UI, no captions."
},
"_meta": {
"title": "Google Gemini"
}
},
"8": {
"class_type": "PreviewAny",
"inputs": {
"source": [
"7",
0
]
},
"_meta": {
"title": "Preview as Text"
}
}
})
# run() also passes api_key to the partner nodes in this workflow.
job = client.run(workflow, api_key=api_key)問答
兩者皆可。已有終格可用 LoadImage 載入,否則在 GeminiNode 輸入描述,並用 GeminiImage2Node 生成符合故事與解析度的終格。
在 ByteDance2FirstLastFrameNode(Seedance 2.0)設定總格數(或片長)與 FPS。格數多或 FPS 高則片長較長、動作更平滑。想要節奏明快可減少格數或略降 FPS。
請確保兩張圖尺寸與比例一致,構圖與光線相近,主題、姿勢、服裝盡量相同。在 Seedance 2.0 降低動作強度或增加片長。若有選項,設定固定種子以重現結果並重跑。
請讓首尾影格尺寸一致。若用 GeminiImage2Node 生成終格,請將其輸出尺寸設為與起始影格相同。自備終格時,請先外部裁切/縮放,確保兩張圖寬高完全一致。
相關工作流程
- LTX-2.3
Turn an image into a video
- MiniMax H3
Create a video from references
- LTX-2.3
Add camera motion first last frame
- Wan2.2
1.2 入門範本-圖像轉影片
- 合作夥伴節點
AI on the Lot 2026 - LTX-2.3 - i2v
- Google Gemini, Kling 3.0, Z-Image-Turbo
AIrt MAchIne(API 版本)
- SSD1.5
AnimateDiff - IPAdapters 迴圈動畫
- Seedance1.0 Pro, ByteDance
ByteDance: 起始結束到影片
- 合合作夥伴節點
FastVideo FastH3: Text to Video
- Flux
FLUX 3 影片:文字轉影片
- Flux
FLUX 3 影片:文字轉影片
- 合作夥伴節點
Gemini Omni 1.1 Flash: Image to Video