demobook

Weavy: Modular character and prompt engineering

Demo summary

Demonstration of building a modular prompting system where individual text boxes for lighting, composition, and subject are concatenated into a final prompt for generation.

Step-by-step

  1. Create individual text boxes for specific prompt categories like lighting, composition, or subject
  2. Label each text box to identify the prompt component
  3. Connect the text boxes to a prompt concatenator node
  4. Input specific descriptors into the boxes to see the final prompt update in real time
  5. Connect the concatenator output to the model nodes for generation

Options

  • Use a system prompt to automatically expand loose ideas into a full prompt
  • Deconstruct character design into categories like ethnicity, gender, age, and body type
  • Combine modular manual inputs with an AI system prompt

Tips

  • Build prompts in a modular fashion to 'art direct' the output more precisely
  • Use a prompt formula to determine which boxes to create
  • Treat modular prompting and AI-assisted prompting as non-mutually exclusive systems that can be used together

Highlights

there's basically no end to how you can do this

All demos from “RoboDojo w/ Rory Flynn - Weavy.ai Masterclass

  1. 35:372:12Building a batch product ad workflow in WeavyRory demonstrates a Weavy workflow that takes a single product image and a user prompt, uses an LLM to generate 10 unique variations, and outputs them as a batch of marketing assets.WeavyAI Product Photography
  2. 40:391:48Scaling e-commerce assets for BarkBoxA demonstration of a Weavy app that preserves product details while generating infinite environmental variations for a BarkBox product, showing the results in a grid view.WeavyAI Product Photography
  3. 53:261:24Connecting nodes for multi-model generationA step-by-step walkthrough of connecting a prompt node to multiple image generation nodes (like Reev and Flux) to produce simultaneous outputs in Weavy.WeavyAI Image Generator
  4. 55:111:02Aesthetic testing across multiple AI modelsRory shows how to route a single prompt to seven different models (including Ideogram, Mystic, and Higgsfield) to compare visual styles side-by-side.WeavyAI Image Generator
  5. 56:472:24Modular character and prompt engineeringCurrentDemonstration of building a modular prompting system where individual text boxes for lighting, composition, and subject are concatenated into a final prompt for generation.WeavyAI Image Generator
  6. 1:07:411:57Multi-model video generation pipelineA demonstration of a video workflow that takes an image and a system prompt to trigger simultaneous video generation across Kling, V3, and SeaArt models.WeavyImage to Video
  7. Watch “RoboDojo w/ Rory Flynn - Weavy.ai Masterclass” →

AI Image Generator

  1. 0:561:06Basic image generation in Weavy AIThe user demonstrates the basic node-based interface of Weavy AI by connecting a prompt node to a Google Imagen 3 node to generate an image of a dog and cat.Artturi Jalli
  2. 3:162:13Building a multi-node image generation workflowThe creator demonstrates building a complex workflow in Weavy AI using prompt concatenators and the Google Imagen 3 model to generate a specific highball glass image.Artturi Jalli
  3. 7:341:48Extracting camera angles using LLM nodesThe user demonstrates resizing a 3D model view and using a ChatGPT node to describe the specific camera angle in text for use in further image generation.Artturi Jalli
  4. 9:311:59Controlled image generation with Flux Depth ProThe creator demonstrates using the Flux Depth Pro node to generate a final high-quality image that preserves the spatial orientation and angle of the previously created 3D model.Artturi Jalli
  5. 53:261:24Connecting nodes for multi-model generationA step-by-step walkthrough of connecting a prompt node to multiple image generation nodes (like Reev and Flux) to produce simultaneous outputs in Weavy.Machine Cinema
  6. 55:111:02Aesthetic testing across multiple AI modelsRory shows how to route a single prompt to seven different models (including Ideogram, Mystic, and Higgsfield) to compare visual styles side-by-side.Machine Cinema
  7. 56:472:24Modular character and prompt engineeringCurrentDemonstration of building a modular prompting system where individual text boxes for lighting, composition, and subject are concatenated into a final prompt for generation.Machine Cinema
  8. 13:492:32Design a YouTube thumbnail with Weavy AIRory Flynn shows how he uses Weavy AI's compositor node and layers to isolate characters, remove backgrounds, and adjust camera angles to create a consistent aesthetic for a podcast thumbnail.Rourke Heath
  9. 21:054:47Build a one-click mood board generatorRory demonstrates a complex Weavy AI workflow that uses Gemini Flash and Flux to batch-process 12 distinct visual prompts from a single image to generate a complete 12-box mood board.Rourke Heath
  10. 34:473:06Scaling image generation with prompt arrays in Weavy AIThe demo shows how to use an LLM node to generate 10 unique prompt variations from a single idea, splitting them via an array node to produce batch outputs in Weavy AI's app mode.Anderson Edjéme
  11. 29:034:45Automating Multi-Angle Prompts with Array and List NodesThe demonstration shows how to use an 'Array' node to split a single block of text (separated by asterisks) into individual camera angle prompts and then mapping them to a 'List' node for batch processing.DGI Kaos
  12. 2:181:57Generate character variations with Flux Pro and Higgsfield in WeavyThe creator demonstrates how to use image generator nodes like Flux Pro and Higgsfield within Weavy, utilizing a text iterator node to run multiple prompts simultaneously for a Cyberpunk character.Sebastian Kamph