Skip to content
All jobs & gigs

Model Evaluation Expert (Audio) (模型评测专家 (音频))

ShengShu Technology (Vidu) · Beijing, China

Full-timePosted Nov 18, 2026

ShengShu Technology (Vidu) is hiring: Model Evaluation Expert (Audio) (模型评测专家 (音频)) (Beijing, China). ShengShu Technology is the Beijing-based company behind Vidu, an AI video-generation model and creation platform. Main focus: For the audio and sound and picture performance in the audio and video generation model, build an evaluation system, continuously evaluate the performance of the internal model in terms of listening, sound quality, sound and picture synchronization and usability, track mainstream products in the industry, and output conclusions that can guide algorithm iteration and product decision-making.

Key responsibilities

  • For the audio and sound and picture performance in the audio and video generation model, build an evaluation system, continuously evaluate the performance of the internal model in terms of listening, sound quality, sound and picture synchronization and usability, track mainstream products in the industry, and output conclusions that can guide algorithm iteration and product decision-making;
  • Evaluation benchmark construction: Design and maintain audio-related evaluation benchmarks for audio and video generation models. Key evaluations include: instruction compliance, sound quality and sense of hearing, timbre consistency, sound and image synchronization, mixing balance and sense of space, as well as professional dimensions such as emotional matching and voice narrative expression;
  • Model evaluation and industry comparison: Based on the evaluation benchmark, carry out internal model iteration testing, and compare with mainstream AIGC products such as Keling and Seedance to analyze the differences in audio effects, integrated sound and picture performance, controllability, stability and production availability;
  • Use cases and problem attribution: Design hierarchical and boundary use cases (dialogue clarity, sound effect event alignment, BGM and emotion matching, noise/distortion, long-term timbre drift, complex sound field, etc.); establish failure mode classification, locate problems such as sound quality degradation, unstable timbre, out of sync, emotion mismatch, instruction deviation, etc., and distinguish model capabilities and prompt/parameter factors;
  • Reporting and decision support: Output evaluation reports with clear structures and reproducible conclusions, give model optimization suggestions, and support version iterations and business decisions;
  • Continuous iteration of the system: Track changes in audio and video generation technology and industry needs, continue to optimize evaluation dimensions, weights and use cases, and improve evaluation efficiency and conclusion credibility.

Requirements / qualifications

  • Bachelor degree or above, majoring in recording art, music production, audio engineering, sound design, digital media and other related majors are preferred. Priority will be given to those with working background in film and television comics/advertising e-commerce/pan-internet;
  • Have practical experience in using AIGC tools such as Keling, Seedance, TapNow, LibTV, etc., and be able to compare the differences in audio and video performance and controllability of different products;
  • Have professional listening and sound aesthetic abilities, and be able to make stable and interpretable judgments on the following dimensions: instruction compliance - whether the sound results accurately respond to prompt words and control conditions; sound quality and listening sense - clarity, noise, distortion, naturalness, etc.; consistency - whether timbre, loudness, and style are stable in timing; sound and picture synchronization - whether dialogue, sound effects, music and picture events are aligned; sound expression - mix balance, sense of space, emotional matching and narrative support
  • At the same time, it has the ability to judge basic picture usability and can participate in audio-visual integrated evaluation;
  • Highly sensitive to production quality, with clear logic and smooth communication, able to independently complete evaluation documents and reports, objective and rigorous, and responsible for evaluation conclusions; fast learner, able to continuously follow up on new products and capabilities;

Nice to have

  • Bonus points: Experience in designing listening standards and evaluation standards, experience in dubbing, sound design, music production, mixing and other related projects, and experience in using AI workflow and performance improvement tools.

Team: R&D Location: Beijing, China Employment type: Full-time Posted date: 2026-08-14 Original title: 模型评测专家 (音频) Original posting is in Chinese; description translated to English.