VPick Image Generator#

Multi-model AI image generation on a visual canvas — Midjourney, Grok, nano-banana, Seedream — with reference image support, style transfer, and batch generation. Powered by VPick.

Security & Data Handling#

• MCP link is a credential: Your MCP config contains an embedded authentication token in the URL ( https://vpick.10xboost.org/mcp/t/xxxxx...). Treat it like a password — do not share it publicly.
• Token scope: The embedded token grants generation access to your VPick account. It can create/manage canvas nodes, trigger image generations, upload reference images, and export results. It cannot access your social media accounts or any services outside VPick.
• Token storage: The token is stored server-side in VPick's database (Google Cloud, us-central1). It is never written to your local filesystem. You can regenerate it anytime at vpick.10xboost.org/settings.
• Data flow: All generation requests go through VPick's server ( vpick.10xboost.org on Google Cloud). VPick routes requests to third-party AI model providers (Midjourney, Grok, Seedream, nano-banana) on your behalf. Your prompts and uploaded reference images are sent to these providers for processing.
• Storage: Generated images are stored in Google Cloud Storage under your VPick account. Uploaded reference images are used only for generation and stored in your project.
• No local credentials: No local API keys, environment variables, or secrets are needed. All auth is embedded in the MCP link.
• Billing: Generation costs are charged to your VPick credit balance, not directly to any third-party service.
• Third-party service: This skill relies on VPick, a cloud AI generation platform. Documentation: vpick-doc.10xboost.org. Source code: github.com/snoopyrain.

Prerequisites#

1. Sign up at vpick.10xboost.org (Google OAuth — new users get $1 free credit)
2. Get your MCP link: Go to Settings → copy your MCP Server URL (contains your auth token)
3. Add to Claude: Paste the MCP link into Claude settings as a Connector — no install, no API keys needed

Available Image Models#

ModelCostOutputBest Fornano-banana-2$0.16/image1 imageDefault, fast, multi-reference supportMidjourney Relaxed$0.045/grid4 images (grid)Budget artistic generationMidjourney Fast$0.12/grid4 images (grid)Quick artistic generationMidjourney Turbo$0.24/grid4 images (grid)Fastest MidjourneyGrok Imagine$0.06/call6 images (T2I) or 2 (I2I)Budget bulk generationSeedream 5.0 Lite$0.0825/image1 image (2K)High resolution, affordableSeedream 5.0 HD$0.0825/image1 image (3K)Ultra high resolution

Aspect Ratios: 1:1, 16:9, 9:16, 3:4, 4:3 (all models)

Core Workflow#

Step 1: Set Up#

Check user's canvas and project:

text

get_canvas          → See current state
list_projects       → Check existing projects
create_project      → Start fresh if needed

Step 2: Create Image Generator Node#

text

add_node(type: "image-generator", name: "My Image")

Step 3: Set the Prompt#

Option A — Direct prompt:

text

update_node(id: "<node_id>", data: { prompt: "a cyberpunk cityscape at sunset, neon lights, rain reflections" })

Option B — Connect a Text node:

text

add_node(type: "text", name: "Prompt", data: { content: "a cyberpunk cityscape at sunset" })
connect_nodes(sourceId: "<text_node>", targetId: "<image_node>", sourceHandle: "text-out", targetHandle: "prompt-in")

Step 4: Add Reference Images (Optional)#

Reference images guide the generation style or subject. Up to 10 reference images supported.

text

upload_image(url: "https://example.com/reference.jpg")
connect_nodes(sourceId: "<upload_node>", targetId: "<image_node>", sourceHandle: "file-out", targetHandle: "image-in")

Use @NodeLabel in prompts to reference connected images:

text

"A portrait in the style of @reference, vibrant colors"

The @reference is resolved at generation time to the connected image URL.

Step 5: Generate#

text

run_image_generator(
  nodeId: "<image_node>",
  prompt: "a cyberpunk cityscape at sunset, neon lights",
  model: "nano-banana-2",
  aspectRatio: "16:9"
)

Midjourney-specific options:

text

run_image_generator(
  nodeId: "<image_node>",
  prompt: "a cyberpunk cityscape at sunset",
  model: "midjourney",
  aspectRatio: "16:9",
  mjSpeed: "fast",        // "relaxed", "fast", "turbo"
  mjVersion: "7.0"        // Midjourney version
)

Grok — auto-detects T2I vs I2I:

• No reference image → Text-to-Image (6 images)
• With reference image → Image-to-Image (2 images)

Step 6: Browse Results#

Models may generate multiple images:

• Midjourney: 4-image grid — browse via currentImageIndex (0-3)
• Grok: 6 images (T2I) or 2 images (I2I)
• Others: 1 image

Check results:

text

get_node(id: "<image_node>")   → See generatedImageUrl, status
list_nodes                      → Overview of all nodes

Step 7: Use Generated Images#

Generated images can be used as input for:

• Video generation: Connect to VideoGenerator's start-image-in port
• More image generation: Connect as reference for another ImageGenerator
• Character consistency: Use as Element reference in MultiShot video

text

connect_nodes(sourceId: "<image_node>", targetId: "<video_node>", sourceHandle: "image-out", targetHandle: "start-image-in")

Common Use Cases#

Character Design (White Background References)#

For video production, generate character references on white backgrounds:

text

run_image_generator(
  nodeId: "<node>",
  prompt: "full body portrait of a young woman, casual outfit, white background, isolated figure, front view",
  model: "nano-banana-2",
  aspectRatio: "3:4"
)

Generate multiple angles (front, side, 3/4) for use as MultiShot Elements.

Batch Generation#

Create multiple ImageGenerator nodes and generate in parallel:

text

add_node(type: "image-generator", name: "Variant A")
add_node(type: "image-generator", name: "Variant B")
add_node(type: "image-generator", name: "Variant C")
run_image_generator(nodeId: "<A>", prompt: "sunset beach, warm tones", model: "midjourney")
run_image_generator(nodeId: "<B>", prompt: "sunset beach, cool tones", model: "grok-imagine")
run_image_generator(nodeId: "<C>", prompt: "sunset beach, dramatic", model: "nano-banana-2")

Style Transfer with References#

text

upload_image(url: "https://example.com/style-reference.jpg")
connect_nodes(sourceId: "<upload>", targetId: "<image_node>", sourceHandle: "file-out", targetHandle: "image-in")
run_image_generator(
  nodeId: "<image_node>",
  prompt: "a portrait in the style of @style-reference, detailed, high quality",
  model: "nano-banana-2"
)

Organize Your Canvas#

Keep things tidy:

text

auto_layout(nodeIds: [...], direction: "horizontal", spacing: 200)
create_group(nodeIds: [...], overrides: { label: "Character Designs", color: "#E57373" })

Check Usage & History#

text

list_generated_files(limit: 20, model: "midjourney")   → Recent generations filtered by model
get_generation_stats                                     → Cost breakdown by model

Tips#

• nano-banana-2 is the best default — fast, affordable, supports multi-reference
• Midjourney excels at artistic/stylized images — use relaxed speed to save cost
• Grok is the cheapest for bulk generation ($0.06 for 6 images)
• Seedream produces the highest resolution (up to 3K)
• White background images are essential for video Element references
• @NodeLabel references are resolved at generation time — make sure labels match exactly
• Use list_models tool to check current pricing and available models

Error Handling#

ErrorSolutionGeneration timeoutAuto-retries up to 2 times; check status with get_nodeInsufficient creditsTop up at vpick.10xboost.orgInvalid aspect ratioUse: 1:1, 16:9, 9:16, 3:4, or 4:3Reference not resolvedEnsure @NodeLabel matches the connected node's label exactlyNode not foundUse list_nodes to get current node IDs

Documentation#

• VPick App: vpick.10xboost.org
• Available models: Call list_models tool for current pricing and capabilities