Paper: Obed Christian Sanjaya, Achmad Zahir Wajdi, Takahiro Tamura, Yoshiaki Shinjo, Yusuke Takahashi, Asako Uraki: “Will It Ship? Production Survivability of Generative AI for Commercial Video,” Proceedings of the International Electronics Symposium 2026 (IES 2026), Yogyakarta, Indonesia, August 1–3, 2026 (Accepted & presented).

Highlighting research from AIx{ }: “Will It Ship?” by Obed Christian Sanjaya, Achmad Zahir Wajdi, Takahiro Tamura, Yoshiaki Shinjo, Yusuke Takahashi, and Asako Uraki — winner of the 🏆 Best Paper Award at IES 2026 in Yogyakarta, Indonesia.

The question isn’t “can it generate?” — it’s “can it survive production?”

Generative AI can now produce realistic visuals from prompts, cheaply. So the operative question for creative work has moved from “can a model make a picture?” to “does it meet commercial production standards — production survivability?” Where the companion position paper “Almost Nobody Can Deliver” argues the case, this paper measures it, on the same campaign.

Eleven models × seven deliverables = a 77-cell matrix

Against a major Japanese personal-care brand’s spring campaign (seven short-form video deliverables), the authors compare eleven generative models (four image, seven video) and score them on a six-dimensional rubric: (i) channel feasibility, (ii) brand fidelity, (iii) anatomical consistency, (iv) physical consistency, (v) human expression, and (vi) regulatory compliance — a 77-cell evaluation matrix in all.

The findings are clear. Still images can be production-grade with reference-anchored prompting. Video requires segmentation into 5–8-second chunks routed to specialized models — never a single pass. And human facial expression stayed below the survivability threshold in every model tested: cuts driven by a face or speech should be shot live.

“Which cut goes to which model”

Figure .1 Production Survivability Routing Matrix. Each of seven short-form video deliverables from a real body-care campaign is scored against five generative tools on a 4-level scale (4 = lead … 1 = unsuitable) by an expert production team using the six-dimension survivability rubric. Read as a routing policy, four deliverables are recommended AI-led and three hybrid (live-action + AI); no single tool wins across every axis. NBP = Nano Banana Pro.

Three contributions: (i) the six-dimension survivability rubric, (ii) an empirical capability map of eleven models, and (iii) a deliverable-to-model routing matrix. Four of the seven deliverables were recommended as AI-led, three as live-action/AI hybrids. The team also built a web companion system that fires one prompt at multiple image models in parallel and scores each with a Vision-LLM judge (Gemini) on the same six dimensions, with a blind expert review offering preliminary validation (human–judge agreement Cohen’s κ = 0.74; inter-rater Fleiss’ κ = 0.71).

The two lead authors (Sanjaya and Wajdi) are Indonesia-based contributors — a fitting international collaboration given the host and organizer (PENS). The throughline: the decisive variable is not raw model capability but the surrounding process — anchors, segmentation, routing, and quality assurance — connecting practitioner know-how to evaluation science. That a paper born from industry–academia and cross-border collaboration went on to earn IES 2026’s Best Paper Award speaks to the strength of its framing and the rigor of its evidence.

Publication details

• Authors: Obed Christian Sanjaya, Achmad Zahir Wajdi, Takahiro Tamura, Yoshiaki Shinjo, Yusuke Takahashi, Asako Uraki (OZU, Inc. / Faculty of Data Science, Musashino University / Asia AI Institute)
• Title: Will It Ship? Production Survivability of Generative AI for Commercial Video
• Venue: International Electronics Symposium 2026 (IES 2026)
• Date & Venue: Yogyakarta, Indonesia, August 1–3, 2026
• Award: 🏆 Best Paper Award — IES 2026
• Status: Accepted & presented
• Web: https://ies.pens.ac.id/2026/

Cite this paper (paste the LaTeX / BibTeX as code blocks — the companion …_Citations.docx has them pre-styled):

IEEE: O. C. Sanjaya, A. Z. Wajdi, T. Tamura, Y. Shinjo, Y. Takahashi, and A. Uraki, “Will It Ship? Production Survivability of Generative AI for Commercial Video,” in Proc. Int. Electron. Symp. (IES), Yogyakarta, Indonesia, Aug. 2026, accepted and presented.

\bibitem{sanjaya2026willitship}
O.~C.~Sanjaya, A.~Z.~Wajdi, T.~Tamura, Y.~Shinjo, Y.~Takahashi, and A.~Uraki, ``Will It Ship? Production Survivability of Generative AI for Commercial Video,'' in \emph{Proc. Int. Electron. Symp. (IES)}, Yogyakarta, Indonesia, Aug.~2026, accepted and presented.

ACM: Obed Christian Sanjaya, Achmad Zahir Wajdi, Takahiro Tamura, Yoshiaki Shinjo, Yusuke Takahashi, and Asako Uraki. 2026. Will It Ship? Production Survivability of Generative AI for Commercial Video. In Proceedings of the International Electronics Symposium 2026 (IES 2026). IEEE, Yogyakarta, Indonesia. Accepted and presented.

\bibitem{sanjaya2026willitship}
Obed Christian Sanjaya, Achmad Zahir Wajdi, Takahiro Tamura, Yoshiaki Shinjo, Yusuke Takahashi, and Asako Uraki. 2026. Will It Ship? Production Survivability of Generative AI for Commercial Video. In \emph{Proceedings of the International Electronics Symposium 2026 (IES 2026)}. IEEE, Yogyakarta, Indonesia. Accepted and presented.

IPSJ (情報処理学会): Sanjaya, O. C., Wajdi, A. Z., Tamura, T., Shinjo, Y., Takahashi, Y. and Uraki, A.: Will It Ship? Production Survivability of Generative AI for Commercial Video, Proceedings of the International Electronics Symposium 2026 (IES 2026) (2026).（accepted and presented）

\bibitem{sanjaya2026willitship}
Sanjaya, O. C., Wajdi, A. Z., Tamura, T., Shinjo, Y., Takahashi, Y. and Uraki, A.: Will It Ship? Production Survivability of Generative AI for Commercial Video, Proceedings of the International Electronics Symposium 2026 (IES 2026) (2026).（accepted and presented）

BibTeX:

@inproceedings{sanjaya2026willitship,
author = {Sanjaya, Obed Christian and Wajdi, Achmad Zahir and Tamura, Takahiro and Shinjo, Yoshiaki and Takahashi, Yusuke and Uraki, Asako},
title = {Will It Ship? Production Survivability of Generative {AI} for Commercial Video},
booktitle = {Proceedings of the International Electronics Symposium 2026 (IES 2026)},
address = {Yogyakarta, Indonesia},
month = aug,
year = {2026},
publisher = {IEEE},
note = {Accepted and presented}
}

This post is part of a series in which AIx{ } — a research project at the Faculty of Data Science, Musashino University — introduces related research.