Research map / Image generation / Text-to-image diffusion
PreciseControl: Enhancing Text-To-Image Diffusion Models with Fine-Grained Attribute Control
Rishubh Parihar, Sachidanand VS, Sabariswaran Mani, Tejan Karmali, Venkatesh Babu RADHAKRISHNAN
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| condition T2I on StyleGAN W+ latent edit direction instead of text prompt for attribute editing (ablation) | PreciseControl · Attribute editing (Beard, Age) | Delta CLIP: Beard 0.986 -> 2.473; Age 0.429 -> 1.777 | Improved |
| use W+ latent editing instead of text-based Imagic editing | PreciseControl · Attribute editing (Beard) | LPIPS: 0.533 -> 0.185 | Improved |
| use W+ latent editing instead of Imagic for identity preservation | PreciseControl · Attribute editing (Beard) | identity similarity (CS): 0.492 -> 0.731 | Improved |
| delay identity token injection until timestep below tau (ablation) | PreciseControl · personalization (qualitative) | text editability | Improved |
| add time-dependent token embeddings (ablation) | PreciseControl latent adaptor · personalization (qualitative) | identity similarity | Improved |
| add subject-specific LoRA tuning (ablation) | PreciseControl · personalization (qualitative) | identity and prompt similarity | Improved |
| add identity loss and regularization loss (ablation) | PreciseControl · personalization (qualitative) | identity/editability | Improved |
| separate subject-specific LoRA models fused with masks instead of joint training (ablation) | multi-subject composition · multi-subject generation (qualitative) | attribute mixing | Improved |
| use text-based Imagic editing vs W+ for edit strength | Imagic · Attribute editing (Smile) | Delta CLIP: Imagic 1.945 vs Ours + W+ 1.104 | Mixed |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- InterfaceGAN — InterFaceGAN: Interpreting the Disentangled Face Representation Learned by GANs (TPAMI 2020) · better on 4 of 8 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- When StyleGAN Meets Stable Diffusion: a W+ Adapter for Personalized Image Generation (CVPR 2024 · similarity 0.91)
- FineStyle: Fine-grained Controllable Style Personalization for Text-to-image Models (NeurIPS 2024 · similarity 0.89)
- Continuous, Subject-Specific Attribute Control in T2I Models by Identifying Semantic Directions (CVPR 2025 · similarity 0.88)
- Text-conditional Attribute Alignment across Latent Spaces for 3D Controllable Face Image Synthesis (CVPR 2024 · similarity 0.88)
- DreamSalon: A Staged Diffusion Framework for Preserving Identity-Context in Editable Face Generation (CVPR 2024 · similarity 0.88)
- Cross Initialization for Face Personalization of Text-to-Image Models (CVPR 2024 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
