International Marketing Internship Report

Toongether.ai: AI-Powered Storytelling Platform

Prepared by: Chad Keith

Student ID: [Student ID]

Institution: California State University, Northridge (CSUN)

Degree Program: Bachelor of Science in Marketing

Internship Period: [Start Date] - [End Date]

Submission Date: October 30, 2023

Executive Summary

This report documents my international marketing internship experience at Toongether.ai, an innovative AI-powered storytelling platform. During my tenure, I focused on enhancing the platform's AI character generation capabilities, with particular emphasis on improving diversity, representation, and technical efficiency.

Key achievements include the recreation of lost art styles, migration of workflows from SDXL to Flux, development of 50 distinct LORAs across various demographics, and implementation of a system capable of generating 950 million unique characters. Most notably, these improvements were accomplished at a fraction of previous costs, reducing compute expenses from approximately $13,000 to just $13.12.

Through an exit interview with CEO Cedric Roux, valuable insights were gained regarding Toongether.ai's target market, business focus, strengths, challenges, and future goals. This internship provided practical application of marketing concepts in an international business context while contributing meaningfully to a company at the intersection of AI technology and creative storytelling.

View our cost efficiency analysis to understand the financial impact of these improvements.

Introduction

As part of the degree requirements for a Bachelor of Science in Marketing at California State University, Northridge (CSUN), I completed an international marketing internship with Toongether.ai. This innovative company operates at the cutting edge of artificial intelligence and creative storytelling, providing a platform that empowers users to create cartoon characters and visual narratives without requiring traditional artistic skills.

The internship presented unique challenges and opportunities, including working across international time zones (requiring 4 AM meetings) and applying marketing principles to an emerging technology space. This report documents the experience, achievements, and insights gained during this period, culminating in an exit interview with CEO Cedric Roux that provided valuable perspective on the company's strategic direction.

About Toongether.ai

Toongether.ai is a consumer-focused platform that leverages artificial intelligence to democratize storytelling and character creation. Through an iOS app, the company empowers novice creators to bring their stories to life without requiring traditional artistic skills or technical knowledge of complex software. The platform emphasizes community building and accessibility, with a mission to lower barriers to creative expression.

Internship Overview

As a Marketing Intern at Toongether.ai, my role extended beyond traditional marketing functions to include technical development that directly impacted the platform's market positioning and user experience. The internship focused on enhancing the AI character generation capabilities, with particular emphasis on improving diversity, representation, and technical efficiency.

Working remotely across international time zones presented both challenges and opportunities. Regular meetings at 4 AM PST with CEO Cedric Roux required significant adaptation but provided valuable experience in international business operations. This arrangement also offered insights into cultural differences in business approaches, particularly regarding monetization strategies and market priorities.

The internship's scope encompassed:

  • Technical development of AI character generation systems
  • Implementation of diversity and representation features
  • Workflow optimization and migration
  • Cost efficiency improvements (see infographic)
  • Market analysis with focus on US cultural preferences

This multifaceted experience provided practical application of marketing concepts while developing technical skills relevant to emerging AI technologies.

Key Accomplishments

Art Style Recreation

Successfully recreated art styles that had been lost when the previous creative director departed. This recovery was critical for maintaining platform consistency and preserving the visual identity that users had come to expect. The recreation process involved analyzing existing assets and developing new training methodologies to replicate the distinctive aesthetic qualities of the original styles.

Workflow Migration: SDXL to Flux

Led the migration of workflows from SDXL to Flux, enabling new capabilities such as direct text integration onto cartoon images. This technical transition laid the groundwork for future feature enhancements and improved the platform's flexibility. The migration required careful planning to ensure continuity while implementing new capabilities that would enhance the storytelling experience.

Diversity & Representation Enhancement

Developed and trained 50 distinct LORAs (Low-Rank Adaptations) covering various demographics: five core ethnicities across different age groups (from babies to seniors) and genders. This initiative significantly improved representation in the platform's character generation capabilities, making it more culturally relevant for the US market where diversity is a key consideration for users.

Cost Efficiency Optimization

Achieved remarkable cost reduction in compute expenses, bringing training costs down from approximately $13,000 to just $13.12 while delivering superior results. This 99.9% cost reduction represents a transformative improvement in operational efficiency, freeing resources for reinvestment in other areas of the platform. View cost efficiency analysis.

Character Generation System Development

Designed and implemented a sophisticated backend system capable of mixing up to three LORAs with different weights, enabling the generation of approximately 950 million unique characters. The system also allows users to save character "recipes," ensuring consistency across different poses and scenes—addressing a key user pain point regarding character continuity.

50
LORAs Developed
950M+
Unique Characters Possible
99.9%
Cost Reduction

Cost Comparison Over Extended Timeframe (Logarithmic Scale)

Company Analysis

Based on the exit interview with CEO Cedric Roux, the following analysis provides insights into Toongether.ai's market positioning, strategic focus, and future direction:

Target Market

Toongether.ai primarily targets novice creators—individuals who have stories to tell but may lack traditional artistic skills or technical knowledge of complex software. The platform focuses exclusively on the consumer market through its iOS app, with no current plans to expand into business or government sectors.

International Strategy

While the company has an international foundation and global availability through app stores, there are no immediate plans for targeted pushes into specific non-English speaking countries with dedicated sales efforts. The current approach relies on organic growth and platform appeal rather than country-specific marketing initiatives.

Company Strengths

According to CEO Cedric Roux, Toongether.ai's greatest strengths include:

  • Accessible AI technology that lowers barriers to creation
  • A passionate community built around storytelling
  • The newly improved, diverse, and cost-effective character generation capability

Challenges & Weaknesses

Key challenges identified include:

  • Limited market awareness and penetration, particularly in large markets like the US
  • Ensuring long-term financial sustainability while maintaining user experience
  • Operational robustness, including asset management and cost control

Future Goals (12-24 Months)

Toongether.ai's primary objectives for the next 1-2 years include:

  • User growth and deepening community engagement
  • Rolling out meaningful feature enhancements (e.g., text-on-image capability)
  • Ensuring platform stability, scalability, and financial sustainability

Marketing Approach

The company does not operate with large, fixed marketing budgets in the traditional sense. Instead, Toongether.ai invests in product development and community support, viewing the product and community as the best marketing channels. This approach prioritizes organic growth and product-led acquisition rather than significant advertising expenditures.

Reflections & Learnings

The internship at Toongether.ai provided valuable insights and learning opportunities that extended beyond traditional marketing education:

"My American brain instantly goes to 'how do we monetize this?'" - This observation during the exit interview highlights cultural differences in business approaches and the value of international exposure in developing a more nuanced understanding of different market philosophies.

Key Learnings:

  1. Technical Marketing Integration: The experience demonstrated how technical development directly impacts marketing positioning and user experience, particularly in AI-driven products.
  2. International Business Operations: Working across significant time zones (4 AM meetings) provided practical experience in global business coordination and communication.
  3. Cultural Business Differences: Exposure to different approaches to monetization and business growth highlighted the importance of cultural context in marketing strategy.
  4. AI Technology Applications: Hands-on experience with cutting-edge AI systems provided valuable technical knowledge that complements traditional marketing education.
  5. Community-Focused Business Models: Observing Toongether.ai's emphasis on community building over aggressive monetization offered insights into alternative business approaches.

The internship successfully bridged academic marketing concepts with practical application in an innovative, international business context. The technical skills developed during this period complement traditional marketing knowledge and provide valuable differentiation in an increasingly technology-driven field.

Conclusion

The marketing internship at Toongether.ai represented a unique opportunity to apply marketing principles in an innovative, international context while contributing meaningfully to a company at the intersection of AI technology and creative storytelling. The experience provided valuable insights into global business operations, cultural differences in business approaches, and the integration of technical development with marketing strategy.

Key accomplishments during the internship period—including the development of 50 LORAs for improved diversity, the creation of a system capable of generating 950 million unique characters, and a 99.9% reduction in compute costs—demonstrate the potential for significant impact even within a time-limited internship role.

The insights gained from CEO Cedric Roux regarding toongether.ai's target market, strengths, challenges, and future direction provide valuable context for understanding the company's position and potential trajectory in the competitive landscape of AI-powered creative tools.

This internship experience has not only fulfilled the requirements for the Marketing degree at CSUN but has also provided practical skills and knowledge that will be valuable in future professional endeavors at the intersection of marketing, technology, and international business.

Cost Efficiency Analysis

Traditional LoRA Training vs. Combinatorial Approach at Scale

$1,752
Traditional Annual Cost
For 400 daily LoRA trainings
$17.52
Combinatorial Annual Cost
Using 50 base LoRAs
99%
Cost Reduction
With combinatorial approach

Traditional Approach

400 LoRA trainings per day

Each training takes 20 minutes on RTX 3090 at 70% power (300W)

Energy Calculation:

300W × (20min/60) × 400 = 40,000 Wh = 40 kWh/day

Cost at $0.12/kWh:

40 kWh × $0.12 = $4.80/day → $146/month → $1,752/year

950M Characters Cost:

$4,140,000 (at 400/day for 950M characters)

Combinatorial Approach

50 Base LoRAs

(5 ethnicities × 5 age groups × 2 genders)

Weighted Combinations

3 LoRAs per generation at weights from 0.05 to 1.00

Energy Calculation:

50 LoRAs × 20min = 16.7h total training → 5 kWh

950M Characters Cost:

Same 50 LoRAs → $0.60 one-time cost

Cost Comparison Over Extended Timeframe (Logarithmic Scale)

Energy Consumption Comparison (Logarithmic Scale)

Toongether.AI Efficiency Report

Key Findings:

  • The combinatorial approach reduces energy consumption by 99.9% compared to traditional LoRA training
  • At 950M characters, traditional training would cost $4.14 million while combinatorial costs $0.60
  • Each additional character in traditional approach adds $0.0044 in cost, while combinatorial adds $0.00000000063
  • The combinatorial method can generate 950M+ unique characters from just 50 base LoRAs
  • Energy savings equivalent to powering 1,400 homes for a year at scale

Implementation Details:

The system uses a random CSV line picker node that selects:

  1. Race from race CSV (5 options)
  2. Age from age CSV (5 options)
  3. Gender from sex CSV (2 options)
  4. Random strength from 0.05 to 1.00 for 3 separate LoRAs

The LoRAs are applied in sequence between the checkpoint output and CLIP input, with model weights controlled by math expression nodes that cycle through weights in 0.05 increments.

Environmental Impact:

At scale, the combinatorial approach could save:

14,600 MWh
Annual energy savings
10,300 tons
CO2 emissions avoided
$4.14M
Annual cost savings at scale

Made with DeepSite LogoDeepSite - 🧬 Remix