PhotoshopAPI 0.8.2

A performant read/write parser of Photoshop Files (*.psd and *.psb)

PhotoshopAPI

[!NOTE] The PhotoshopAPI is still in early development status which means it is subject to change and will likely include bugs. If you find any please report them to the issues page

About

PhotoshopAPI is a C++20 Library with Python bindings for reading and writing of Photoshop Files (*.psd and *.psb) based on previous works from psd_sdk, pytoshop and psd-tools. As well as the official Photoshop File Format Specification, where applicable. The library is continuously tested for correctness in its core functionality. If you do find a bug please submit an issue to the github page.

The motivation to create another library despite all the other works present is that there isn't a library which has layer editing as a first class citizen while also supporting all bit-depths known to Photoshop (8-bits, 16-bits, 32-bits). This Library aims to create an abstraction between the raw binary file format and the structure that the user interfaces against to provide a more intuitive approach to the editing of Photoshop Files.

Why should you care?

Photoshop itself is unfortunately often slow to read/write files and the built-in tools for automatically/programmatically modifying files suffer this same issue. On top of this, due to the extensive history of the Photoshop File Format, Photoshop files written out by Photoshop itself are often unnecessarily bloated to add backwards compatibility or cross-software compatibility.

The PhotoshopAPI tries to address these issue by allowing the user to read/write/modify Photoshop Files without ever having to enter Photoshop itself which additionally means, no license is required. It is roughly 5-10x faster in reads and 20x faster in writes than photoshop while producing files that are consistently 20-50% lower in size (see the benchmarks section on readthedocs for details). The cost of parsing is paid up front either on read or on write so modifying the layer structure itself is almost instantaneous (except for adding new layers).

Features

Supported:

• Read and write of *.psd and *.psb files
• Creating and modifying simple and complex nested layer structures
• Smart Objects (replacing, warping, extracting)
• Pixel Masks
• Modifying layer attributes (name, blend mode etc.)
• Setting the Display ICC Profile
• 8-, 16- and 32-bit files
• RGB, CMYK and Grayscale color modes
• All compression modes known to Photoshop

Planned:

• Support for Adjustment Layers
• Support for Vector Masks
• Support for Text Layers
• Indexed, Duotone Color Modes

Not Supported:

• Files written by the PhotoshopAPI do not contain a valid merged image in order to save size meaning they will not behave properly when opened in third party apps requiring these (such as Lightroom)
• Lab and Multichannel Color Modes

Python

The PhotoshopAPI comes with Python bindings which can be installed using

$ py -m pip install PhotoshopAPI

alternatively the wheels can be downloaded from the Releases page. For examples on how to use the python bindings please refer to the Python Bindings section on Readthedocs or check out the PhotoshopExamples/ directory on the github page which includes examples for Python as well as C++.

For an even quicker way of getting started check out the Quickstart section!

Documentation

The full documentation with benchmarks, build instructions and code reference is hosted on the PhotoshopAPI readthedocs page.

Requirements

This goes over requirements for usage, for development requirements please visit the docs.

• A CPU with AVX2 support (this is most CPUs after 2014) will greatly increase performance, if we detect this to not be there we disable this optimization
• A 64-bit system
• C++ Library: Linux, Windows or MacOS
• Python Library¹: Linux, Windows, MacOS

The python bindings support python >=3.7 (except for ARM-based MacOS machines which raise this to >=3.10)

¹ Currently Linux is supported only as manylinux build and has some features disabled such as timestamps on logging.

Performance

The PhotoshopAPI is built with performance as one of its foremost concerns. Using it should enable you to optimize your pipeline rather than slow it down. It runs fully multithreaded with SIMD instructions to leverage all the computing power your computer can afford.

As the feature set increases this will keep being one of the key requirements. For detailed benchmarks please visit the docs

Below you can find some of the benchmarks comparing the PhotoshopAPI ('PSAPI') against Photoshop in read/write performance

Quickstart

The primary struct to familiarize yourself with when using the PhotoshopAPI is the LayeredFile as well as all its Layer derivatives (such as ImageLayer and GroupLayer), all of these are template structs for each of the available bit depths.

To get a feel of what is possible with the API as well as how to use it please refer to PhotoshopExample/ directory. To familiarize yourself with the main concepts, as well as recommended workflows check out the docs or the examples.

If more fine grained control over the binary structure is necessary, one can modify the PhotoshopFile which is what is parsed by the API internally. Do keep in mind that this requires a deep understanding of how the Photoshop File Format works.

Below is a minimal example to get started with opening a PhotoshopFile, removing some layer, and writing the file back out to disk:

C++

using namespace NAMESPACE_PSAPI;

// Initialize some constants that we will need throughout the program
const static uint32_t width = 64u;
const static uint32_t height = 64u;

// Create an 8-bit LayeredFile as our starting point, 8- 16- and 32-bit are fully supported
LayeredFile<bpp8_t> document = { Enum::ColorMode::RGB, width, height };
// Create our individual channels to add to our image layer. Keep in mind that all these 3 channels need to 
// be specified for RGB mode
std::unordered_map <Enum::ChannelID, std::vector<bpp8_t>> channelMap;
channelMap[Enum::ChannelID::Red] = std::vector<bpp8_t>(width * height, 255u);
channelMap[Enum::ChannelID::Green] = std::vector<bpp8_t>(width * height, 0u);
channelMap[Enum::ChannelID::Blue] = std::vector<bpp8_t>(width * height, 0u);

ImageLayer<bpp8_t>::Params layerParams = {};
layerParams.name = "Layer Red";
layerParams.width = width;
layerParams.height = height;

auto layer = std::make_shared<ImageLayer<bpp8_t>>(std::move(channelMap), layerParams);
document.add_layer(layer);

// It is perfectly legal to modify a layers properties even after it was added to the document as attributes
// are only finalized on export
layer->opacity(.5f);

// Convert to PhotoshopFile and write to disk. Note that from this point onwards 
// our LayeredFile instance is no longer usable
LayeredFile<bpp8_t>::write(std::move(document), "WriteSimpleFile.psd");

The same code for reading and writing can also be used to for example LayeredFile::move_layer_ or LayeredFile::add_layer as well as extracting any image data

Python

import os
import numpy as np
import photoshopapi as psapi

# Initialize some constants that we will need throughout the program
width = 64
height = 64
color_mode = psapi.enum.ColorMode.rgb

# Generate our LayeredFile instance
document = psapi.LayeredFile_8bit(color_mode, width, height)

img_data = np.zeros((3, height, width), np.uint8)
img_data[0] = 255
# When creating an image layer the width and height parameter are required if its not a zero sized layer
img_layer = psapi.ImageLayer_8bit(img_data, "Layer Red", width=width, height=height)
document.add_layer(img_layer)

# Similar to the C++ version we can adjust parameters of the layer after it has been added to the document
# as long as it happens before we write to disk
img_layer.opacity = .5

document.write(os.path.join(os.path.dirname(__file__), "WriteSimpleFile.psd"))