historIX

Turning Textbooks into Talk: Interactive Learning with Historical Figures Combining generative AI video synthesis and conversational models, historiX redefines history learning by turning passive content into active dialogue with the past.

DELIVERABLES

IOS mobile app GenAI Images + Video including voice

Year

2025

Role

Creative Direction

Challenges

  • Old fashioned teaching methods are faced with challenges of being passive and hard to connect with, leading to low engagement and retention of students.


  • As a result, teachers are faced with challenges of making historical content relatable within limited time.

  • Current tools often don’t match the digital fluency or expectations of Gen Z students


  • Historical figures feel distant and inaccessible - Students rarely empathize with or deeply understand the people behind the events.

Challenges

  • Old fashioned teaching methods are faced with challenges of being passive and hard to connect with, leading to low engagement and retention of students.


  • As a result, teachers are faced with challenges of making historical content relatable within limited time.

  • Current tools often don’t match the digital fluency or expectations of Gen Z students


  • Historical figures feel distant and inaccessible - Students rarely empathize with or deeply understand the people behind the events.

Solution

  • An application where high school students can choose a historical figure to learn about, and the system will create a 3D version of them, including replicating their voice

  • The student can choose their questions in real time to start the live simulation.


  • The application can be accessed through a mobile or their tablet device.


Process

  • Led both product and marketing design as BRIA’s sole designer during early growth phase.

  • Tested and validated design usability with early adopters


  • Introduced a more streamlined workflow from concept to dev handoff, cutting design-to-build time.


Solution

  • An application where high school students can choose a historical figure to learn about, and the system will create a 3D version of them, including replicating their voice

  • The student can choose their questions in real time to start the live simulation.


  • The application can be accessed through a mobile or their tablet device.


Key Takeaways

  • User Testing with multiple AI Voice and sound software:
    Tested multiple AI tools to generate 3D historical figures from text, image, video, and voice. Voice cloning (using ElevenLabs) was the most challenging, especially for lesser-known figures, with limited audio references like I tested, on Rabbi Joseph B. Soloveitchik.


  • He had an Eastern European accent mixed with American as he lived several years in New York. I wanted to create an answer to my question for him based on how he would most likely respond and in an accent as close to his as possible.  

    ​

    • I downloaded snippets of his voice online

    • I then uploaded it to Elevenlabs and synced it with a polish accent of an older man to try get his accent.


    • I Improved output by editing background noise for cleaner, more accurate replication.




  • User testing with multiple visual (video) editing software:


    I tested different prompts on leonardo AI and trained models to try to get it as close as possible to the original historical figure

    ​

    MUST HAVE KEY PROMPTS:

    • Front Facing Angle of camera should be facing camera to get a higher quality image

    • Realistic quality I tested skin textures and used prompts of skin textures and pores to show human like skin quality 

    • Test on different models Some models work better than others for realistic  human photography. I tested and trained different models until I got a result that were most accurate.


  • User Testing for image using the model 4.0 of chatGPT and video with lip-sync using Runway AI:
    I tested different lip syncing models to match the text I wanted to create the final step.


  • Visual generation on chatGPT Vs. Leonardo AI: Prompts were far less detailed on chatGPT and in some cases, more accurate,  on well known figures, where there is presumably more visual data available online.​



Key Takeaways

  • User Testing with multiple AI Voice and sound software:
    Tested multiple AI tools to generate 3D historical figures from text, image, video, and voice. Voice cloning (using ElevenLabs) was the most challenging, especially for lesser-known figures, with limited audio references like I tested, on Rabbi Joseph B. Soloveitchik.


  • He had an Eastern European accent mixed with American as he lived several years in New York. I wanted to create an answer to my question for him based on how he would most likely respond and in an accent as close to his as possible.  

    ​

    • I downloaded snippets of his voice online

    • I then uploaded it to Elevenlabs and synced it with a polish accent of an older man to try get his accent.


    • I Improved output by editing background noise for cleaner, more accurate replication.




  • User testing with multiple visual (video) editing software:


    I tested different prompts on leonardo AI and trained models to try to get it as close as possible to the original historical figure

    ​

    MUST HAVE KEY PROMPTS:

    • Front Facing Angle of camera should be facing camera to get a higher quality image

    • Realistic quality I tested skin textures and used prompts of skin textures and pores to show human like skin quality 

    • Test on different models Some models work better than others for realistic  human photography. I tested and trained different models until I got a result that were most accurate.


  • User Testing for image using the model 4.0 of chatGPT and video with lip-sync using Runway AI:
    I tested different lip syncing models to match the text I wanted to create the final step.


  • Visual generation on chatGPT Vs. Leonardo AI: Prompts were far less detailed on chatGPT and in some cases, more accurate,  on well known figures, where there is presumably more visual data available online.​



Key Takeaways

  • User Testing with multiple AI Voice and sound software:
    Tested multiple AI tools to generate 3D historical figures from text, image, video, and voice. Voice cloning (using ElevenLabs) was the most challenging, especially for lesser-known figures, with limited audio references like I tested, on Rabbi Joseph B. Soloveitchik.


  • He had an Eastern European accent mixed with American as he lived several years in New York. I wanted to create an answer to my question for him based on how he would most likely respond and in an accent as close to his as possible.  

    ​

    • I downloaded snippets of his voice online

    • I then uploaded it to Elevenlabs and synced it with a polish accent of an older man to try get his accent.


    • I Improved output by editing background noise for cleaner, more accurate replication.




  • User testing with multiple visual (video) editing software:


    I tested different prompts on leonardo AI and trained models to try to get it as close as possible to the original historical figure

    ​

    MUST HAVE KEY PROMPTS:

    • Front Facing Angle of camera should be facing camera to get a higher quality image

    • Realistic quality I tested skin textures and used prompts of skin textures and pores to show human like skin quality 

    • Test on different models Some models work better than others for realistic  human photography. I tested and trained different models until I got a result that were most accurate.


  • User Testing for image using the model 4.0 of chatGPT and video with lip-sync using Runway AI:
    I tested different lip syncing models to match the text I wanted to create the final step.


  • Visual generation on chatGPT Vs. Leonardo AI: Prompts were far less detailed on chatGPT and in some cases, more accurate,  on well known figures, where there is presumably more visual data available online.​



Next Project

Next Project

2025 ® NICOLE SCHATTNER

2025 ® NICOLE SCHATTNER

2025 ® NICOLE SCHATTNER