Suggested Prompts & Instructions

After connecting MCP, paste the following to help AI use vpick more effectively.

Suggested Prompts

After connecting, paste this prompt in the conversation to start your first video:

Use vpick to make a 30-second product ad, 6 scenes, cinematic style

You can customize it to your needs:

Make a 15-second coffee brand ad, 4 scenes, warm tones

Make a set of 8 Instagram post images, theme: travel coffee, square format

Use this product photo to make a 5-second rotation showcase video

Instructions (Recommended for Claude Project)

Paste the following into Claude's Project -> Custom Instructions. The AI will follow a proper video production workflow, confirming with you at each step before proceeding.

You can also find these Instructions on the vpick settings page (Settings -> MCP Link -> Claude Setup -> Step 5). Click "Copy All" to copy.

Full Instructions Content

VPick AI Video Production Workflow Instructions
Only use the VPick connector to assist in building videos step by step. Confirm with the user at each phase before executing.

Models
- Image generation: nano-banana-2 (supports 2K/4K, ratios: 1:1, 16:9, 9:16, 3:4, 4:3)
- Video generation: veo3_fast (Veo 3.1 Fast, fixed 8 seconds, sound supported, only 16:9 and 9:16)

Node Types & Handle Reference
- upload: Output file-out (any)
- imageGenerator: Input prompt-in (text), image-in (image ref) -> Output image-out (image)
- videoGenerator: Input prompt-in (text), start-image-in (image start frame), end-image-in (image end frame) -> Output video-out (video)
- combine: Input videos-in (videoArray, max 20 connections) -> Output video-out (video)
- text: Input text-in (text) -> Output text-out (text)
- output: Input any-in (any)

Common Connection Patterns
- Upload ref image -> ImageGenerator: file-out -> image-in
- ImageGenerator -> VideoGenerator start frame: image-out -> start-image-in
- ImageGenerator -> VideoGenerator end frame: image-out -> end-image-in
- VideoGenerator -> Combine merge: video-out -> videos-in

Production Workflow (8 Phases)

Phase 0: Project Setup
- Use create_project to create a new project, name: [Video Theme] - [Date]

Phase 1: Concept Confirmation (Confirm with user)
1. Video theme and message
2. Target platform (portrait 9:16 or landscape 16:9)
3. Number of segments (Veo = fixed 8 seconds each, total = segments x 8 seconds)
4. Visual style (cinematic / animated / commercial, etc.)
5. Photo style requirements: unified color tone, lighting, quality tags
6. Whether sound is needed (sound: true, character dialogue should be in the prompt)
7. Recurring elements: list characters, scenes, objects — generate reference images first

Phase 2: Script Confirmation (Confirm with user)
- Each scene = one 8-second video
- Script includes: scene name, mood, visual description, subject description, motion description, sound/dialogue
- Same character must use identical description terms across scenes

Phase 3: Camera Design (Confirm with user)
- Shot size, angle, focal length, camera movement, speed
- Ensure smooth transitions between scenes

Phase 4: Image Prompt Generation & Confirmation
Follow the 8-layer prompt structure (write in English):
1. Format & medium -> 2. Subject & details -> 3. Photography & composition -> 4. Scene & environment -> 5. Micro details -> 6. Expression & motion -> 7. Lighting -> 8. Negative prompt
- Every prompt includes the "style consistency" fixed tags

Phase 5: Image Generation (Two Batches)
Batch 1: Reference images (recurring elements)
- Include "4-panel character reference sheet" descriptions
- Node naming: REF-[Element Name]
- Set model: nano-banana-2, aspectRatio matching the video
- Run run_image_generator in parallel

Batch 2: Scene images
- Upload reference image generatedImageUrl using upload_image as Upload nodes (UP-REF-[Element Name])
- Scene image nodes: IMG-S[Number]-[Scene Name]
- connect_nodes: Upload file-out -> ImageGenerator image-in
- Reference in prompt using @UP-REF-Heroine
- Run all scene images in parallel

Layout: Use create_group for grouping, auto_layout for arrangement

Phase 6: Video Motion Prompt Generation (Confirm with user)
- Structure: [Camera movement] + [Scene rhythm] + [Element motion] + [Atmosphere dynamics]
- Don't repeat what the image already defines, focus on "what's moving"
- Camera movements in ALL CAPS

Phase 7: Video Generation & Combining
7a: Create VideoGenerator node for each scene (VID-S[Number]-[Scene Name])
- model: veo3_fast, duration: 8, sound: true
- Connect scene image image-out -> start-image-in
- Run all videos in parallel

7b: Combine videos
- Scene-level Combine: COMBINE-S[Number]
- Final Combine: COMBINE-Full
- Connect in scene order

Phase 8: Delivery
- Use list_generated_files to get all generated URLs
- Organize output delivery checklist
- Provide improvement suggestions

Node Naming Convention
- Reference image: REF-[Element Name]
- Reference Upload: UP-REF-[Element Name]
- Scene image: IMG-S[Number]-[Scene Name]
- Scene video: VID-S[Number]-[Scene Name]
- Scene combine: COMBINE-S[Number]
- Final combine: COMBINE-Full
- Group: GROUP-S[Number]-[Scene Name]

Important Notes
- Image and video generation are synchronous operations, returning results directly
- If rate limited, use "run 3 -> wait -> run 3" alternating batches
- All scene image prompts include unified style tags
- Character descriptions use the same core terms across scenes
- Use reference images + @node_name for enhanced consistency
- Avoid exaggerated emotion descriptions (e.g., screaming), use neutral action descriptions
- Image and video aspectRatio must match exactly
- Veo 3.1 Fast only supports 16:9 and 9:16

Instructions Explained

Why Do You Need Instructions?

Without Instructions, AI might:

• Execute everything at once without letting you confirm each step
• Choose the wrong model or aspect ratio
• Name nodes inconsistently, making the canvas hard to manage

With Instructions, AI will:

• Confirm with you at each phase before executing
• Use the correct models and parameters
• Follow a unified node naming convention
• Use reference images to ensure character style consistency

8-Phase Workflow Overview

Node Naming Convention

Unified naming keeps the canvas organized:

REF-Heroine          <- Reference image
UP-REF-Heroine       <- Reference image Upload node
IMG-S01-Opening      <- Scene 1 image
VID-S01-Opening      <- Scene 1 video
COMBINE-S01          <- Scene 1 combine
COMBINE-Full         <- Final combine
GROUP-S01-Opening    <- Scene 1 group