TECNO, the AI-driven, technology brand, revealed at MWC 2026 its Edge-Side AIGC Preview Concept Technology, developed in collaboration with Arm, the industry’s highest-performing and most power-efficient compute platform.
The partnership brings real-time, on-device generative AI to mobile devices and creative applications, while delivering significant efficiency gains that help scale these advanced experiences across various device tiers.
As generative AI moves from novelty to everyday utility, mobile access is being shaped by practical constraints, such as connectivity, latency and device processing power. Creative workloads are compute-intensive, many experiences rely on continuous internet access, and even minor latency can disrupt the user experience while photo editing or framing. Sustained AI processing also impacts device performance.
Edge AI is a key pillar of TECNO’s broader AI strategy. TECNO has slimmed down complex algorithms and large models, improving efficiency through tailored solutions built on the Arm® compute platform. The result is real-time preview generation at 30fps, a split-screen interface making outputs more tangibly felt, and overall system performance improvements. Moreover, the preview experience utilises 100% on-device AI, ensuring consistency regardless of network connectivity.
“Generative AI creativity only matters to users if the feedback is instant. We’re optimising the on-device pipeline so creators can preview, iterate, and produce without waiting. This enables a real-time, immersive experience that delivers better user satisfaction, stronger compatibility across device tiers, and ultimately makes TECNO’s already-advanced imaging technologies even more compelling and intuitive,” said Minjie Chen, PhD, Head of Imaging Algorithms Department, TECNO.
Real-Time, Split-Screen Preview at 30fps
TECNO undertook deep, low-level code construction and introduced a customised Style Transfer Preview technology that uses algorithmic optimisation to reduce screen flickering and jitters commonly seen in traditional solutions. By migrating the inference pipeline to the Arm CPU, the Style Transfer Preview is real-time at 30fps and ensures a fluid split-screen viewfinder experience. As users move the camera, the viewfinder updates instantly, enabling immediate preview and confident composition without delay.
Model efficiency is another key focus. TECNO has compressed models to run fully offline on smartphones while preserving artistic fidelity. This model compression enhances practical utilisation of the Arm CPU for compute-intensive tasks such as image restoration, ensuring stable performance ready for everyday use.
The Edge-Side AIGC Preview Concept Technology leverages advanced Armv9 CPU capabilities, seamlessly accessed through Arm Kleidi™AI. By leveraging the Armv9 CPU’s specialised i8MM (Int8 Matrix Multiplication) instructions through KleidiAI, the collaboration accelerates core machine learning (ML) operations across mainstream frameworks. This synergy delivers 30% faster and more efficient AI compute, providing the robust foundation necessary for real-time, low-latency creative experiences.
“As generative AI moves from the cloud to the edge, the smartphone becomes the most important AI device in people’s lives,” said Steve Raphael, Senior Director, Smartphone Segment, Edge AI at Arm. “Armv9 CPUs deliver the sustained performance and efficiency required for real-time, on-device AI. Together with TECNO, we’re showing that advanced creative experiences can run fully offline and at scale – bringing immersive, instant creativity to users everywhere.”
As the industry shifts from cloud-first AI to edge-first creativity, TECNO is advancing on-device generative AI to reduce friction such as latency, buffering, and inconsistent performance. Through ongoing collaboration with Arm, the company aims to bring generative AI experiences – once confined to workstations – directly onto smartphones through on-device optimisation and platform-level efficiency, making next‑generation creativity accessible to more users.