Elena Cavallin - AI Projects & Research

GenAI projects & research

Projects, tools, workshops, and research focused on integrating GenAI in creative processes. Custom chatbots, educational experiences, and practical applications in design workflows.

Projects

Concept Smart Glasses

Maintained visual consistency across product contexts using ChatGPT prompt strategies.

Tools and workshops

Weavy AI Video Generation

Weavy

For visual consistency in multi-channel projects. I generate style-matched video/image sequences from existing brand guidelines to maintain coherence between motion graphics and static designs.

Firefly Design

Firefly

Adobe Firefly integrated in Photoshop and Illustrator for production-ready edits: generative fill, background extension, and style-consistent variations.

Kling AI Creation

Kling

Rapid prototyping of motion concepts and animated storyboards. Generate multiple animation variations to test timing and movement before final production.

Runway ML Workshop

Runway

Transform static concepts into dynamic narratives. I convert Midjourney outputs into video sequences and apply motion effects. Learned during Monogrid Workshop (December 2023).

Midjourney Visual Design

Midjourney

Visual exploration and concept development. Systematic approach with iterations, custom style codes, and blend functions. Trained during "visiting week 2023" at UNIRSM.

Lexica Interior Design

Lexica

Early AI experimentation tool. Used with Iuav students (2022-2023) across interior and product design programs to develop workflows for coherent image generation.

Custom chatbots

BIAS Chatbot

BIAS

Identifies cognitive biases in environmental decisions through targeted questions. Helps develop awareness of prejudices that influence climate change projects.

Try BIAS

IAmleto GPT

IAmleto

AI co-author for scriptwriting. Guides writers from initial research to final draft of theatrical works, screenplays, and radio dramas.

Try IAmleto

speculative chatbot

Speculative assistant

Converts speculative research into visual concepts. Analyzes project files and produces AI prompts for materializing future scenarios.

Try it

SOSIA GPT

SOSIA

Selects sustainable materials based on application type and project requirements. Filters options by technical and environmental criteria.

Try SOSIA

WAVe 2025 - Connected Visions

Week-long workshop (June 23-27, 2025) at Università Iuav di Venezia with ANFAO. Students designed connected eyewear concepts as interfaces between body and technology.

Combined traditional design thinking with generative AI tools to create visual narratives for wearable technology concepts.

Tool Selection

Selected Kling based on favorable credit and workflow compatibility. Suitable for translating static concept boards into motion sequences.

Workflow

Pipeline integrating ChatGPT for co-prompting, Midjourney for static imagery, Kling for final video sequences. Style consistency from ideation to motion.

Students learned structured prompt patterns: role framing, design intent, constraints, and style systems.

Custom Support

Created specialized assistants: "Speculative Assistant" (ChatGPT Plus users) and "Speculation Assistant" (free users) for effective prompt engineering.

Structured templates adapted to different access levels, ensuring quality outputs regardless of subscription status.

Process

Iterative cycles: think → prompt → generate → critique → refine.

Results

Students developed critical awareness of AI capabilities, limitations, and bias considerations. They produced static postcards illustrating speculative futures and a video summarizing the scenario.

Monogrid Workshop - December 2023

Hands-on training on AI-powered video synthesis and motion design. Covered video generation capabilities from text-to-video to advanced image-to-video transformations.

Focus on practical applications in design workflows and professional project integration.

My Approach

Start with Midjourney-generated static concepts, feed them into Runway for motion sequences. Maintains visual consistency while adding dynamic elements.

Combine video synthesis with AI-generated audio to create complete audiovisual narratives for client presentations.

Static to motion transformation example

Advanced motion graphics example

Applications

Rapid prototyping of motion concepts and animated storyboards. Generate multiple animation variations quickly to test timing and movement.

Reduces iteration time in pre-visualization, enabling faster decision-making during client presentations.

Integration

Use when clients need to see how static designs work in motion. Quickly demonstrate transitions, hover effects, and animated sequences.

AI outputs serve as reference material for final production, providing clear direction for motion designers.

Advanced motion graphics example

Process

Generate multiple motion variations from single static input, analyze movement patterns, select most effective approaches for project goals.

Use Cases

Brand campaigns requiring both static and motion assets. Consistency between AI-generated static images and animated counterparts ensures cohesive visual communication.

Visiting week 2023

Two-day workshop at UNIRSM University of San Marino, organized by Maria Luce Lupetti (TU Delft, now Politecnico di Torino).

Introduced 49 second-year Web Design students to GenAI technologies with focus on Midjourney for professional visual content.

Methods

Integrated ChatGPT for structured prompts and Midjourney for image generation. Students learned advanced prompt engineering including "blend" function for image fusion.

Emphasized creative iteration, documented entire process through Discord to analyze design strategy evolution.

Activities

Day 1: GenAI introduction, Midjourney familiarization through self-portraits, future scenarios, web assets (hero images, logos).

Day 2: Job advertisement portraits, diversity and inclusivity focus, product visualization for brand identity.

Outcomes

Students developed Midjourney-specific vocabulary, learned advanced parameter control, and managed creative blocks through prompt reformulation.

Analysis revealed skill evolution patterns from basic usage to sophisticated stylistic controls.

Hero images generated for web projects