Block 1
Visual Seed Image
Create the approved image reference that anchors the video scene.
Model
@promptha.recraft-v4-pro
Preset
dynamic-visual-seed
Output
visualSeedImage · image
Create a cinematic warm morning story shot of a small neighborhood bakery opening before sunrise. The storefront glows with amber window light, a baker unlocks the door, steam rises from fresh bread trays, and the street is quiet and blue before daylight. Generate one polished visual seed image first, then turn it into a 9 second Seedance video with a slow push-in, subtle steam, warm reflections, and a clean final hold.
Generate one polished visual seed image, accept it, then animate it into a controlled Seedance video scene.
Ready
2/2 blocks
Output
Seedance
The settings below show how each block in this example was configured: model route, core inputs, preset settings, prompt, and quality checks.
Block 1
Create the approved image reference that anchors the video scene.
Model
@promptha.recraft-v4-pro
Preset
dynamic-visual-seed
Output
visualSeedImage · image
These are the visual or media outputs from the run. Text-only, memory, and JSON support nodes stay in the workflow map below.

visualSeedImage
Create the approved image reference that anchors the video scene.
seedanceSceneVideo
Animate the accepted visual seed into one practical story video shot.
Accepted visual seed image and generated Seedance video packaged as one reusable guide result.
visualSeedImage
The guide is a reusable production system. Each node creates one accepted output, and that accepted output becomes the source material for the next node.
This run started with: Create a cinematic warm morning story shot of a small neighborhood bakery opening before sunrise. The storefront glows with amber window light, a baker unlocks the door, steam rises from fresh bread trays, and the street is quiet and blue before daylight. Generate one polished visual seed image first, then turn it into a 9 second Seedance video with a slow push-in, subtle steam, warm reflections, and a clean final hold.
Users can generate, revise, and compare candidates, but only the accepted version is passed forward.
Examples shown here must include a visual, carousel, image, or video output. Text-only and JSON-only runs are filtered out.
The last nodes collect accepted outputs into a package that can be saved, reused, or sent to render services.
This is the actual order of the guide. Each output key is the contract the next node can read.
Create the approved image reference that anchors the video scene.
Example goal
Create a cinematic warm morning story shot of a small neighborhood bakery opening before sunrise. The storefront glows with amber window light, a baker unlocks the door, steam rises from fresh bread trays, and the street is quiet and blue before daylight. Generate one polished visual seed image first, then turn it into a 9 second Seedance video with a slow push-in, subtle steam, warm reflections, and a clean final hold.
User customization
Describe the story moment, subject, setting, visual style, mood, camera framing, lighting, and what the later video should preserve.
Command
/image
Model
@promptha.recraft-v4-pro
Preset
dynamic-visual-seed
Single video-ready frame
Image has one clear subject, setting, composition, and motion potential.
No generated text artifacts
Avoids readable text, fake logos, and watermark-like marks.
/image @promptha.recraft-v4-pro visualSeedImage: Generate exactly one polished visual seed image for a downstream video generation block. Use the user goal as the story moment. Preserve one clear subject, one clear setting, readable visual hierarchy, cinematic composition, and a video-friendly frame. Avoid readable text, logos, watermarks, distorted anatomy, cluttered backgrounds, and extra subjects unless requested. Return one image artifact only and store it as visualSeedImage. Guide: Visual Seed to Seedance Video Example goal: Create a cinematic warm morning story shot of a small neighborhood bakery opening before sunrise. The storefront glows with amber window light, a baker unlocks the door, steam rises from fresh bread trays, and the street is quiet and blue before daylight. Generate one polished visual seed image first, then turn it into a 9 second Seedance video with a slow push-in, subtle steam, warm reflections, and a clean final hold. Return output for: Visual seed image. Store it as visualSeedImage. Customization field shown to users: Describe the story moment, subject, setting, visual style, mood, camera framing, lighting, and what the later video should preserve.
Block 2
Animate the accepted visual seed into one practical story video shot.
Model
@promptha.seedance-2.0
Preset
seedance-practical-scene
Output
seedanceSceneVideo · video
Source image
/block-samples/seedance-scene-video-source-field.png
What should happen?
Animate a golden wheat field into a controlled cinematic scene: slow push-in, warm sunlight shimmer, wheat heads swaying in a light breeze, and a stable final hold.
Per-shot prompts
Shot 1, 0-2s: begin from the accepted source frame and start a slow push-in. Shot 2, 2-4s: wheat heads sway naturally while dust motes and sun flare move softly. Shot 3, 4-5s: settle into a clean warm field hold.
What must stay locked?
Preserve the field angle, golden wheat, soil foreground, sun position, warm color, and 16:9 composition.
Upstream accepted outputs
visual-seed-image
Duration
5
Shot plan
three-shot-sequence
Camera
slow-push-in
Format
16:9
Motion intensity
controlled
Ending frame
clean-hold
Uses accepted source image
Video follows the selected visualSeedImage as source truth.
Clean final hold
Clip ends with a usable final frame and avoids chaotic motion.
Source image required
The block must receive a source image URL before video generation.
Single video output
The block must return one video URL for seedanceSceneVideo.
/video @promptha.seedance-2.0 seedanceSceneVideo: Generate exactly one Seedance 2.0 video render. Shot structure: - Total duration: 5 seconds. - Aspect ratio: 16:9. - Shot plan: three-shot-sequence. - Each shot must have its own prompt with duration, camera, subject action, environment motion, transition, and final-frame intent. Source image URL: /block-samples/seedance-scene-video-source-field.png Scene objective: Animate a golden wheat field into a controlled cinematic scene: slow push-in, warm sunlight shimmer, wheat heads swaying in a light breeze, and a stable final hold. Per-shot prompt plan: Shot 1, 0-2s: begin from the accepted source frame and start a slow push-in. Shot 2, 2-4s: wheat heads sway naturally while dust motes and sun flare move softly. Shot 3, 4-5s: settle into a clean warm field hold. Shot setup: - Camera move: slow-push-in. - Motion intensity: controlled. - Ending frame: clean-hold. Preserve: Preserve the field angle, golden wheat, soil foreground, sun position, warm color, and 16:9 composition. Creative references: @style/warm-documentary Seedance 2.0 production direction: - Treat the source image as the first frame and the continuity anchor. - State the shot count and timing before describing visuals. - If no per-shot prompt plan is provided, derive a concise plan from the scene objective and keep it simple enough for one render. - Preserve the subject identity, product shape, wardrobe, location, lighting logic, composition, and visual style unless the user explicitly asks to change them. - Add useful motion that serves each shot: camera movement, human micro-motion, atmosphere, light shifts, fabric movement, reflections, steam, or object movement. - Keep transitions intentional and easy to follow; do not make a random montage. - Keep motion physically believable and easy to render. Avoid chaotic camera movement. - Keep the final frame useful for a cut to the next shot. - Do not add readable text, logos, watermarks, fake UI, deformed anatomy, duplicate subjects, or random new objects. - Extra avoid rules: [Negative Rules] - Return one video artifact only and store it as seedanceSceneVideo. Guide: Visual Seed to Seedance Video Example goal: Create a cinematic warm morning story shot of a small neighborhood bakery opening before sunrise. The storefront glows with amber window light, a baker unlocks the door, steam rises from fresh bread trays, and the street is quiet and blue before daylight. Generate one polished visual seed image first, then turn it into a 9 second Seedance video with a slow push-in, subtle steam, warm reflections, and a clean final hold. Accepted upstream artifacts: - visual-seed-image Return output for: Seedance scene video. Store it as seedanceSceneVideo. Customization field shown to users: Example: make it 9 seconds, vertical, slower push-in, more atmospheric steam, stronger final hold, or stricter subject preservation.
Create the approved image reference that anchors the video scene.
seedanceSceneVideo
Animate the accepted visual seed into one practical story video shot.
Accepted as
visualSeedImage
Feeds into
Seedance Video Scene
Animate the accepted visual seed into one practical story video shot.
Accepted as
seedanceSceneVideo
Feeds into
Final package
Preserve source continuity
The generated video should preserve the source subject, scene, and visual style.
Useful ending frame
The final frame should be stable enough for downstream guide composition or another shot.
Shot structure declared
Seedance 2.0 prompts should state shot count, timing, and per-shot direction before the visual detail.