Periodic Reporting for period 2 - POWDER (Powder is the only app offering a complete turnkey solution to automatically detect the best gaming moments, earn rewards from quests and challenges, and compete with other gamers.)
Reporting period: 2023-08-01 to 2024-07-31
Project context and objectives
The project by Powder aims to address market needs in the gaming industry for creators, creator programs, and gaming brands. It focuses on cost savings and revenue generation through features like highlight extraction, montage creation, file management, rights clearance, audience growth, monetization, paid marketing, and user acquisition.
Project pathway to impact
Powder for Creators: Built from +1800 creators' surveyed, automating highlight extraction, montage creation, and optimizing social media posts. Reducing the 5 hours 17 minutes typically spent on editing per stream.
Powder for Creator Programs: Streamlines content management that invests in highlight extraction, providing enhanced file management and rights clearance, expediting workflows and enhancing team collaboration.
Powder Paid Marketing: Fresh monetization avenues for the top 100 streamers and beyond, compensated views on vertical video, mitigating the need for 1.1 billion views/year for substantial earnings.
The Gaming Destination Platform: Powder into a comprehensive gaming hub for the 1,000-100,000 followers range creators, unlocking additional revenue streams based on Perf. Marketing, targeting the $10bn PC games marketing market.
Scale and significance of expected impacts
The project targets the global gaming market, home to 7 million streamers and millions of brands, set to revolutionize the industry. Powder aims to increase streamer productivity by reducing the 5 hours 17 minutes spent on editing. This approach can aid half of the streamers who hire professional help for content creation, resulting in significant cost savings (+1800 people surveyed). Our ambition is to onboard creators on our platform by leveraging performance marketing for the gaming influencer industry, we foresee unlocking up to $5bn in revenue, particularly for smaller creators with high engagement rates.
In conclusion, Powder promises significant cost savings, increased revenue generation for creators, agencies, and gaming brands, and a significant shift in the gaming industry's landscape. We aim to deliver transformative results and substantial impact, redefining the value of gaming content creation and its monetization.
The project by Powder aims to address market needs in the gaming industry for creators, creator programs, and gaming brands. It focuses on cost savings and revenue generation through features like highlight extraction, montage creation, file management, rights clearance, audience growth, monetization, paid marketing, and user acquisition.
Project pathway to impact
Powder for Creators: Built from +1800 creators' surveyed, automating highlight extraction, montage creation, and optimizing social media posts. Reducing the 5 hours 17 minutes typically spent on editing per stream.
Powder for Creator Programs: Streamlines content management that invests in highlight extraction, providing enhanced file management and rights clearance, expediting workflows and enhancing team collaboration.
Powder Paid Marketing: Fresh monetization avenues for the top 100 streamers and beyond, compensated views on vertical video, mitigating the need for 1.1 billion views/year for substantial earnings.
The Gaming Destination Platform: Powder into a comprehensive gaming hub for the 1,000-100,000 followers range creators, unlocking additional revenue streams based on Perf. Marketing, targeting the $10bn PC games marketing market.
Scale and significance of expected impacts
The project targets the global gaming market, home to 7 million streamers and millions of brands, set to revolutionize the industry. Powder aims to increase streamer productivity by reducing the 5 hours 17 minutes spent on editing. This approach can aid half of the streamers who hire professional help for content creation, resulting in significant cost savings (+1800 people surveyed). Our ambition is to onboard creators on our platform by leveraging performance marketing for the gaming influencer industry, we foresee unlocking up to $5bn in revenue, particularly for smaller creators with high engagement rates.
In conclusion, Powder promises significant cost savings, increased revenue generation for creators, agencies, and gaming brands, and a significant shift in the gaming industry's landscape. We aim to deliver transformative results and substantial impact, redefining the value of gaming content creation and its monetization.
Powder is developing an artificial intelligence system specifically designed for recognizing highlights in video games through the construction of three models:
AI model capable of recognizing video games being played using supervised learning.
AI models that identify game events in a game using supervised learning, 40+ models.
AI audio model that identifies highlights in a video game by anonymously analyzing players' emotions.
Anonymous emotion detections on facial expression have been granted a patent, Publication of US20220189076A1.
We collected and processed large datasets of video game footage to train the models using supervised learning techniques. We implemented and fine-tuned the algorithms, optimizing them for average precision and low resource usage. Additionally, we conducted experiments and evaluations to assess the effectiveness and efficiency of the models in recognizing and extracting game highlights.
The main achievements include:
Development of a robust AI model: Capable of recognizing and classifying over 80 video games with an accuracy exceeding 90%.
Supervised Leaning Innovation: For identifying highlights in video games. Designing a hierarchical classification layer on top of a small image encoder, and adding a post-processing layer allowing to interpret events produced by the image encoder running on sampled frames from the video.
Anonymously recognition of game highlight: analyzing players' emotions via audio.
Implementation and optimization of the models: Ensuring average precision and low resource usage in real-time scenarios.
We generate 2 million AI-generated clips per month and content is also shared across social medias.
Articles associated with our research:
Mariano Rodríguez, Liam Schoneveld, Vincent Garcia. Robust validation steps for clip classification. 2022 ⟨hal-03579068⟩.
Hacene Terbouche, Maryan Morel, Mariano Rodriguez, Alice Othmani; Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, 2023, pp. 3142-3151.
H. Terbouche, L. Schoneveld, O. Benson and A. Othmani, "Comparing Learning Methodologies for Self-Supervised Audio-Visual Representation Learning," in IEEE Access, vol. 10, pp. 41622-41638, 2022, doi: 10.1109/ACCESS.2022.3164745.
AI model capable of recognizing video games being played using supervised learning.
AI models that identify game events in a game using supervised learning, 40+ models.
AI audio model that identifies highlights in a video game by anonymously analyzing players' emotions.
Anonymous emotion detections on facial expression have been granted a patent, Publication of US20220189076A1.
We collected and processed large datasets of video game footage to train the models using supervised learning techniques. We implemented and fine-tuned the algorithms, optimizing them for average precision and low resource usage. Additionally, we conducted experiments and evaluations to assess the effectiveness and efficiency of the models in recognizing and extracting game highlights.
The main achievements include:
Development of a robust AI model: Capable of recognizing and classifying over 80 video games with an accuracy exceeding 90%.
Supervised Leaning Innovation: For identifying highlights in video games. Designing a hierarchical classification layer on top of a small image encoder, and adding a post-processing layer allowing to interpret events produced by the image encoder running on sampled frames from the video.
Anonymously recognition of game highlight: analyzing players' emotions via audio.
Implementation and optimization of the models: Ensuring average precision and low resource usage in real-time scenarios.
We generate 2 million AI-generated clips per month and content is also shared across social medias.
Articles associated with our research:
Mariano Rodríguez, Liam Schoneveld, Vincent Garcia. Robust validation steps for clip classification. 2022 ⟨hal-03579068⟩.
Hacene Terbouche, Maryan Morel, Mariano Rodriguez, Alice Othmani; Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, 2023, pp. 3142-3151.
H. Terbouche, L. Schoneveld, O. Benson and A. Othmani, "Comparing Learning Methodologies for Self-Supervised Audio-Visual Representation Learning," in IEEE Access, vol. 10, pp. 41622-41638, 2022, doi: 10.1109/ACCESS.2022.3164745.
Powder project delivers cutting-edge results in event detection and pattern recognition for gaming, extracting real-time data with minimal user impact. It employs video detection, audio analysis, and compact model development to extract 17 highlights per VOD in under 10 minutes for every 2 hours of stream.
From January 2023, it's averaged 2 million AI-generated clips monthly, boosting user engagement from 10% to 27% stickiness. Long-term success necessitates substantial finance for R&D, infrastructure, marketing, expansion, and team operations.
Estimated budget breakdown:
R&D: €10m for enhancing AI models, integration with new games/platforms, and developing features.
Technical infrastructure: €5m for scalable servers, data storage, and optimized architecture.
Marketing and promotion: €8m for targeted campaigns, partnerships, and tournaments/events.
International expansion: €7m for localization, global infrastructure, and distribution agreements.
Team and operations: €10m for expansion, project management, product development, and operational costs.
The estimated financial and capital needs for market entry are around €40 million, in addition to the €14.2 million already invested. These investments support technology development, market penetration, and an exceptional user experience, ensuring long-term success.
From January 2023, it's averaged 2 million AI-generated clips monthly, boosting user engagement from 10% to 27% stickiness. Long-term success necessitates substantial finance for R&D, infrastructure, marketing, expansion, and team operations.
Estimated budget breakdown:
R&D: €10m for enhancing AI models, integration with new games/platforms, and developing features.
Technical infrastructure: €5m for scalable servers, data storage, and optimized architecture.
Marketing and promotion: €8m for targeted campaigns, partnerships, and tournaments/events.
International expansion: €7m for localization, global infrastructure, and distribution agreements.
Team and operations: €10m for expansion, project management, product development, and operational costs.
The estimated financial and capital needs for market entry are around €40 million, in addition to the €14.2 million already invested. These investments support technology development, market penetration, and an exceptional user experience, ensuring long-term success.