Jinyang Zhang
Generative AI researcher.
I focus on digital humans, image restoration, multimodal generation, and efficient model inference. I am currently seeking full-time research and engineering opportunities.
Recent Project
All projects
Style-Talking
A zero-shot audio-driven digital human system that builds speaking-style prompts from a reference video, predicts LivePortrait expression motion from wav2vec audio features, and renders talking videos while preserving the original non-face regions.
Voice Studio
A local-first macOS voice creation app for premium voice generation, voice cloning, voice design, dialogue generation, and DeepSeek-assisted dialogue rewrite. It targets novels, audio drama, narration, character dialogue, and multi-role audio export workflows.
IConFace
IConFace retains same-identity references as dense visual evidence and reinforces target-aligned degraded structure, recovering persistent localized identity details while also supporting blind face restoration.
Recent Publications
All publicationsIConFace: Fine-Grained Identity Conditioning for Reference-Aware Face Restoration
arXiv preprint arXiv:2605.02814
A fine-grained identity-conditioned framework that retains dense reference evidence and reinforces target-aligned degraded structure.
The Second Challenge on Real-World Face Restoration at NTIRE 2026: Methods and Results
arXiv preprint arXiv:2604.10532v2 / CVPR 2026 Workshop
NTIRE 2026 real-world face restoration challenge report covering task design, weighted IQA evaluation, identity checking, submitted methods, and final rankings.
Learning Discriminative Compact Representation for Hyperspectral Imagery Classification
IEEE Transactions on Geoscience and Remote Sensing (TGRS)
A multi-task deep network for hyperspectral image classification that jointly learns compact spectral representation, reconstruction, and pixelwise classification.
Foundations
All foundations
LLM 核心机制交互讲义
从 BPE 到 next token 的理论、交互实验与可运行代码
AIGC 与 LLM 数学基础
系统整理 AIGC、LLM、Diffusion、RLHF 与 DPO 相关数学概念
大模型算法岗常见面试题深度解析
Transformer、训练、微调、推理、RAG、Agent 与评估的结构化问答