OmniPSD: Unified PSD Generation & Decomposition

• OmniPSD is a unified framework for generating and decomposing layered PSD files, leveraging an RGBA-VAE and diffusion transformer to achieve high-fidelity design synthesis.
• It supports both text-to-PSD and image-to-PSD tasks by encoding RGBA layers into a shared latent space, ensuring semantically coherent and hierarchically organized outputs.
• The framework demonstrates superior performance with low error metrics and high similarity scores, validated through comprehensive evaluations and user studies.

OmniPSD is a unified generative and decomposition framework for layered PSD (Photoshop Document) files, which supports both text-to-PSD generation and image-to-PSD decomposition while preserving semantically coherent and hierarchically structured RGBA layers. Built upon the Flux ecosystem and leveraging a DiT-style diffusion transformer in an RGBA-VAE latent space, OmniPSD establishes a new paradigm in editable design synthesis, enabling high-fidelity, transparency-aware layered designs from textual prompts or single input images (Liu et al., 10 Dec 2025).

1. Problem Formulation: Dual Generative and Decomposition Tasks

OmniPSD addresses two core tasks, both within a shared latent space induced by an RGBA-VAE encoder $E_\alpha$ and decoder $D_\alpha$:

• Text→PSD Generation: The input is a hierarchical prompt $$\mathbf{c} = \{\text{caption}_\text{full}, \text{caption}_\text{fg}, \text{caption}_\text{mid}, \text{caption}_\text{bg}\}$$. The output is $L=4$ RGBA layers $$\{I^{(l)}\in\mathbb{R}^{H\times W\times 4}\}_{l=1}^4$$, corresponding to full poster, foreground, midground, and background. These are visually arranged as

$$G = \begin{bmatrix} I^{(1)} & I^{(2)} \ I^{(3)} & I^{(4)} \end{bmatrix} \in \mathbb{R}^{2H\times 2W\times 4}$$

The grid $G$ is encoded via $E_\alpha$ into latent tokens $\mathbf{z}_0$, and a diffusion/flow transformer is trained to map noise to this latent grid, i.e., learning $p_\theta(\mathbf{z}_0|\mathbf{c})$.

• Image→PSD Decomposition: The input is a single flattened RGBA poster $D_\alpha$0. The output is a stack of layers $D_\alpha$1. Each flattened input and target layer is encoded: $D_\alpha$2, $D_\alpha$3. A flow-matching vector field

$D_\alpha$4

is learned for layerwise extraction and erasure via iterative LoRA-adapters.

This dual problem formulation enables both end-to-end design synthesis from descriptions and full layer decomposition for direct PSD editability.

2. Diffusion-Transformer Architecture and Attention Mechanisms

OmniPSD utilizes a two-stage DiT-style transformer operating over RGBA-VAE latent tokens:

• Token Embedding: The RGBA-VAE encodes each image into a spatial grid $D_\alpha$5 of tokens.
• Multi-head Self-attention: Given token matrix $D_\alpha$6:

$D_\alpha$7

with outputs concatenated and projected.

• Cross-attention: For conditional generation, queries are drawn from the diffusion stream, and keys/values from condition tokens (text or flattened images):

$D_\alpha$8

• Spatial In-context Reasoning: By arranging RGBA layers into a $D_\alpha$9 grid, self-attention connects across all layers (e.g., foreground↔background, foreground↔full poster), increasing compositional capacity without specialized modules.

This configuration allows the same transformer backbone to serve both generation and decomposition in a unified architecture.

3. Diffusion and Flow-matching Objectives

OmniPSD employs distinct objectives depending on task:

• DDPM-style Diffusion (Text→PSD):
  • Forward noising:

  $$\mathbf{c} = \{\text{caption}_\text{full}, \text{caption}_\text{fg}, \text{caption}_\text{mid}, \text{caption}_\text{bg}\}$$0

  with $$\mathbf{c} = \{\text{caption}_\text{full}, \text{caption}_\text{fg}, \text{caption}_\text{mid}, \text{caption}_\text{bg}\}$$1. - Reverse denoising by the conditional DiT:

  $$\mathbf{c} = \{\text{caption}_\text{full}, \text{caption}_\text{fg}, \text{caption}_\text{mid}, \text{caption}_\text{bg}\}$$2 - Loss: Simplified $$\mathbf{c} = \{\text{caption}_\text{full}, \text{caption}_\text{fg}, \text{caption}_\text{mid}, \text{caption}_\text{bg}\}$$3 parameterization:

  $$\mathbf{c} = \{\text{caption}_\text{full}, \text{caption}_\text{fg}, \text{caption}_\text{mid}, \text{caption}_\text{bg}\}$$4

• ODE Flow-matching (Image→PSD):

  • Continuous trajectory: $$\mathbf{c} = \{\text{caption}_\text{full}, \text{caption}_\text{fg}, \text{caption}_\text{mid}, \text{caption}_\text{bg}\}$$5
  • Vector field $$\mathbf{c} = \{\text{caption}_\text{full}, \text{caption}_\text{fg}, \text{caption}_\text{mid}, \text{caption}_\text{bg}\}$$6 is learned to match:

  $$\mathbf{c} = \{\text{caption}_\text{full}, \text{caption}_\text{fg}, \text{caption}_\text{mid}, \text{caption}_\text{bg}\}$$7 - This flow-matching approach is deterministic, enabling faster inference in decomposition.

The framework incorporates LoRA-adapters for task specialization in the iterative Image→PSD extraction/erasure pipeline.

4. RGBA-VAE Representation: Transparency Preservation

The RGBA-VAE, an “AlphaVAE” extension, is tailored to editable design with four-channel (RGBA) support:

• Encoder $$\mathbf{c} = \{\text{caption}_\text{full}, \text{caption}_\text{fg}, \text{caption}_\text{mid}, \text{caption}_\text{bg}\}$$8:

$$\mathbf{c} = \{\text{caption}_\text{full}, \text{caption}_\text{fg}, \text{caption}_\text{mid}, \text{caption}_\text{bg}\}$$9

• Decoder $L=4$0: decodes latent representations of color and alpha channels to reconstruct the original image.

• Loss Function:

$L=4$1

with $L=4$2, $L=4$3 representing patch and perceptual feature extractors; $L=4$4 an isotropic Gaussian prior.

Retraining on layered-poster data yields high PSNR/SSIM and low LPIPS, outperforming alternative Alpha-VAEs (MSE=$L=4$5, PSNR=32.5 dB, SSIM=0.945) (Liu et al., 10 Dec 2025).

5. In-context Grid and Iterative Editing Algorithms

OmniPSD’s in-context learning leverages a $L=4$6 grid structure for compositional reasoning during both tasks:

• Text→PSD, Single-pass:

1. Form a $L=4$7 grid $L=4$8 from the four target RGBA layers.
2. Encode $L=4$9 with RGBA-VAE; apply diffusion or flow-matching process in latent space.
3. Decode the recovered latent tokens to obtain all four RGBA output layers.

• Image→PSD, Multi-step Iterative Extraction (pseudocode excerpt):

$$\{I^{(l)}\in\mathbb{R}^{H\times W\times 4}\}_{l=1}^4$$4 Here, each $$\{I^{(l)}\in\mathbb{R}^{H\times W\times 4}\}_{l=1}^4$$0 is a LoRA adapter trained via $$\{I^{(l)}\in\mathbb{R}^{H\times W\times 4}\}_{l=1}^4$$1 on Flux-Kontext, ensuring reversible foreground/background separation.

This grid-based approach enables the transformer to reason over layer compositionality and semantic relationships directly within the attention mechanism.

6. Dataset and Empirical Evaluation

• Layered Poster Dataset: Approximately 200k professionally designed PSDs, each split into RGBA text, foreground, and background layers. For evaluation, 500 hierarchical text prompts and 500 flattened test images are utilized.

• Metrics:

| Task | Metric | OmniPSD | GPT-Image-1 | LayerDiffuse | |--------------|------------------|----------|-------------|--------------| | Text→PSD | FID | 30.43 | 53.21 | 89.35 | | Text→PSD | CLIP-Score (%) | 37.64 | 35.59 | 24.78 | | Text→PSD | GPT-4 Score | 0.90 | 0.84 | 0.66 | | Image→PSD | MSE | 1.14e-3 | 2.48e-2 | N/A | | Image→PSD | PSNR (dB) | 24.0 | N/A | N/A | | Image→PSD | SSIM | 0.952 | N/A | N/A | | Image→PSD | GPT-4 Score | 0.92 | 0.86 | 0.84 |

• Ablation Studies:
  • RGBA-VAE: MSE=$$\{I^{(l)}\in\mathbb{R}^{H\times W\times 4}\}_{l=1}^4$$2, PSNR=32.5 dB, SSIM=0.945 vs. Alpha-VAE: $$\{I^{(l)}\in\mathbb{R}^{H\times W\times 4}\}_{l=1}^4$$3/26.9dB/0.739.
  • Hierarchical prompts: Removal increases Text→PSD FID to 38.56 and reduces the GPT-4 score to 0.78.

User studies (18 participants) confirm the semantic coherence and practical editability of OmniPSD-generated and -decomposed layers. This suggests that transparency-aware latent representations and in-context transformer architectures are effective for editable design generation and decomposition at scale (Liu et al., 10 Dec 2025).