Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

PLACID: Identity-Preserving Multi-Object Compositing via Video Diffusion with Synthetic Trajectories

Gemma Canet Tarrés, Manel Baradad, Francesc Moreno-Noguer, Yumeng Li

TL;DR

PLACID addresses the challenge of studio-quality multi-object compositing by leveraging a pretrained image-to-video diffusion model with text guidance and a synthetic trajectory training pipeline. It introduces high‑resolution object conditioning and object-aware textual tokens to preserve identity while closely matching background colors and layouts. A diverse training data generator—comprising In-the-Wild images, manual designs, and Subject-200k derived data plus synthetic side-by-side scenes—equips the model to learn temporally coherent object perturbations from random initial placements to a final caption-guided composition. Quantitative metrics and user studies show PLACID outperforms prior methods in object identity, background fidelity, and color accuracy, with broader capabilities including editing and video generation; the approach offers a practical, scalable solution for professional multi-object compositing.

Abstract

Recent advances in generative AI have dramatically improved photorealistic image synthesis, yet they fall short for studio-level multi-object compositing. This task demands simultaneous (i) near-perfect preservation of each item's identity, (ii) precise background and color fidelity, (iii) layout and design elements control, and (iv) complete, appealing displays showcasing all objects. However, current state-of-the-art models often alter object details, omit or duplicate objects, and produce layouts with incorrect relative sizing or inconsistent item presentations. To bridge this gap, we introduce PLACID, a framework that transforms a collection of object images into an appealing multi-object composite. Our approach makes two main contributions. First, we leverage a pretrained image-to-video (I2V) diffusion model with text control to preserve objects consistency, identities, and background details by exploiting temporal priors from videos. Second, we propose a novel data curation strategy that generates synthetic sequences where randomly placed objects smoothly move to their target positions. This synthetic data aligns with the video model's temporal priors during training. At inference, objects initialized at random positions consistently converge into coherent layouts guided by text, with the final frame serving as the composite image. Extensive quantitative evaluations and user studies demonstrate that PLACID surpasses state-of-the-art methods in multi-object compositing, achieving superior identity, background, and color preservation, with less omitted objects and visually appealing results.

PLACID: Identity-Preserving Multi-Object Compositing via Video Diffusion with Synthetic Trajectories

TL;DR

PLACID addresses the challenge of studio-quality multi-object compositing by leveraging a pretrained image-to-video diffusion model with text guidance and a synthetic trajectory training pipeline. It introduces high‑resolution object conditioning and object-aware textual tokens to preserve identity while closely matching background colors and layouts. A diverse training data generator—comprising In-the-Wild images, manual designs, and Subject-200k derived data plus synthetic side-by-side scenes—equips the model to learn temporally coherent object perturbations from random initial placements to a final caption-guided composition. Quantitative metrics and user studies show PLACID outperforms prior methods in object identity, background fidelity, and color accuracy, with broader capabilities including editing and video generation; the approach offers a practical, scalable solution for professional multi-object compositing.

Abstract

Recent advances in generative AI have dramatically improved photorealistic image synthesis, yet they fall short for studio-level multi-object compositing. This task demands simultaneous (i) near-perfect preservation of each item's identity, (ii) precise background and color fidelity, (iii) layout and design elements control, and (iv) complete, appealing displays showcasing all objects. However, current state-of-the-art models often alter object details, omit or duplicate objects, and produce layouts with incorrect relative sizing or inconsistent item presentations. To bridge this gap, we introduce PLACID, a framework that transforms a collection of object images into an appealing multi-object composite. Our approach makes two main contributions. First, we leverage a pretrained image-to-video (I2V) diffusion model with text control to preserve objects consistency, identities, and background details by exploiting temporal priors from videos. Second, we propose a novel data curation strategy that generates synthetic sequences where randomly placed objects smoothly move to their target positions. This synthetic data aligns with the video model's temporal priors during training. At inference, objects initialized at random positions consistently converge into coherent layouts guided by text, with the final frame serving as the composite image. Extensive quantitative evaluations and user studies demonstrate that PLACID surpasses state-of-the-art methods in multi-object compositing, achieving superior identity, background, and color preservation, with less omitted objects and visually appealing results.
Paper Structure (34 sections, 1 equation, 21 figures, 3 tables)

This paper contains 34 sections, 1 equation, 21 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Methodology
  4. Model Architecture
  5. Training Strategy
  6. Training Data Generation
  7. Experiments
  8. Evaluation Dataset.
  9. Comparison to State of the Art
  10. Ablation Studies
  11. Training Data.
  12. Emerging Capabilities
  13. Limitations
  14. Conclusions
  15. Emerging Capabilities
  16. ...and 19 more sections

Figures (21)

  • Figure 1: Our model enables simultaneous compositing of multiple objects into seamless natural scenes. Generation can optionally be guided by descriptive captions (top), including specific color instructions (top left). The model ensures all specified objects are present (top middle) and allows them to interact with the given background through specific instructions (top right) or creative compositions (bottom).
  • Figure 2: Limitations of comanici2025gemini in color accuracy and object consistency. Models often (i) alter colors, and (ii) omit, duplicate, or modify objects when combining multiple items.
  • Figure 3: Model Architecture. Our model conditions a DiT video model on images and text. The visual inputs, encoded via radford2021learningCLIP, include: (i) first frame $F_1$ (a random assembly of unprocessed object images), (ii) individual object images $I_{1..N}$, and (iii) an optional background $B$. A caption $c$ describing the desired composition is encoded via raffel2020exploringT5. Image and text encodings are fed to the DiT through separate cross-attention mechanisms. The model flexibly handles varying numbers of objects, with or without $B$.
  • Figure 4: Training Data Generation. Top: Naive way of generating training video data, by simply interpolating the first and last frame. Bottom: Proposed method of generating motion-based temporal consistent video data, in which objects follow synthetic trajectories from initial to final position. Final frame from unsplash.
  • Figure 5: Comparison to State of the Art. We compare our multi-object compositing model to VACE jiang2025vace, UNO wu2025uno, DSD cai2025dsd, OmniGen xiao2025omnigen, MS-Diffusion wang2024ms, NanoBanana comanici2025gemini and Qwen-Image-Edit wu2025qwen.
  • ...and 16 more figures