About This Role
This role involves building and refining the professional evaluation system for the Seedance 2.5 video generation model. You will assess the model's output from a specialized perspective, such as directing or screenwriting, to identify issues affecting the credibility and usability of the generated content. The position suits an expert in film, drama, or related creative fields with a strong interest in generative AI and video models.
Responsibilities
- Establish a professional evaluation system for the Seedance 2.5 video generation model, defining evaluation dimensions, grading standards, typical examples, and acceptance processes around capabilities such as narrative, camera work, scenes, characters, actions, styling, and sound-picture relationships.
- Review model results from a professional perspective (e.g., director, screenwriter, art, photography, performance, action, or costume), identify specific issues affecting the credibility and creative usability of the final product, and break down judgments like "incorrect visuals" or "unconvincing performance" into clear, reproducible problem descriptions.
- Design evaluation tasks and stress test sets covering single-shot, multi-shot narrative, multi-character interaction, complex movements, spatiotemporal continuity, and style consistency, to verify the model's capability boundaries in real creative workflows.
- Establish model version comparison and problem attribution mechanisms, outputting quantitative results, typical cases, and cause breakdowns, distinguishing influencing factors such as prompts, data, models, and product pipelines, and promoting the implementation of optimization priorities and retesting plans.
- Consolidate professional aesthetics and industry experience into executable evaluation specifications, annotation guidelines, case libraries, and calibration mechanisms to improve judgment consistency among different evaluators and support large-scale evaluation.
- Delve into customer scenarios in advertising, film, short drama, animation, gaming, e-commerce, etc., converting customer creative goals and feedback into evaluation tasks to ensure evaluation conclusions reflect actual production effects, and collaborate with algorithm, product, data, and creator teams to participate in data strategy, capability definition, and version acceptance, continuously tracking domestic and international generative video products, professional content standards, and cutting-edge evaluation methods to iterate the Seedance 2.5 evaluation system.
Requirements
- Bachelor's degree or above, with a preference for majors in drama and film directing, drama and film literature, film studies, art, photography, performance, dance, costume design, animation, or related design evaluation.
- Possess solid fundamental skills and mature judgment standards in at least one of the aforementioned professional areas, with project experience in film, TV series, short dramas, advertising, animation, gaming, stage, or fashion, and able to provide representative works or project cases.
- Able to translate subjective creative judgments into clear evaluation dimensions, grading standards, positive and negative examples, and improvement suggestions, with accurate judgment on problem boundaries and priorities.
- Strong interest in generative AI and video models, willing to frequently use Seedance and mainstream image and video generation tools, and understand the impact of prompts, reference images, character assets, and multi-round generation on results.
- Excellent documentation, case analysis, and cross-team communication skills, able to align evaluation conclusions with product, algorithm, data, and creative teams and drive problem resolution.
- Stable aesthetic judgment, specific feedback, maintaining learning speed for new models and workflows, and able to adapt to parallel evaluation of multiple versions.
Nice to have:
- Experience in generative video, multimodal models, or large model evaluation, having participated in the construction of evaluation sets, annotation specifications, version acceptance, or red team testing.
- Project experience in film, animation, gaming, advertising, stage, or fashion, with works that have been launched, shortlisted for professional festivals/exhibitions, or won industry awards.
- Familiar with data annotation, Benchmark design, statistical analysis, or evaluation toolchains, able to participate in the construction of large-scale, automated evaluation solutions.
- Project experience in AI-native short films, virtual production, digital humans, interactive narratives, or complex AIGC content.
About the Company
Volcano Ark is a one-stop large model service platform launched by Volcano Engine, leading the product and market share in China's large model market. The platform provides full-process services including model inference, evaluation, and fine-tuning. Ark integrates Doubao and mainstream industry large models, offers a rich plugin ecosystem and AI application development services, and ensures the implementation of enterprise-level AI applications through stable and reliable secure mutual trust solutions and professional algorithm technical services.
