The code in imgui.cpp embeds a copy of 'ProggyClean.ttf' (by Tristan Grimmer),
a 13 pixels high, pixel-perfect font used by default. We embed it in the source code so you can use Dear ImGui without any file system access. ProggyClean does not scale smoothly, therefore it is recommended that you load your own file when using Dear ImGui in an application aiming to look nice and wanting to support multiple resolutions.
You need to load a font with explicit glyph ranges if you want to use non-ASCII characters. See [Fonts Loading Instructions](#fonts-loading-instructions). Use [Debug Tools](#debug-tools) confirm loaded fonts and loaded glyph ranges.
This is a current constraint of Dear ImGui (which we will lift in the future): when loading a font you need to specify which characters glyphs to load.
All loaded fonts glyphs are rendered into a single texture atlas ahead of time. Calling either of `io.Fonts->GetTexDataAsAlpha8()`, `io.Fonts->GetTexDataAsRGBA32()` or `io.Fonts->Build()` will build the atlas. This is generally called by the Renderer backend, e.g. `ImGui_ImplDX11_NewFrame()` calls it. **If you use custom glyphs ranges, make sure the array is persistent** and available during the calls to `GetTexDataAsAlpha8()/GetTexDataAsRGBA32()/Build()`.
This is often of byproduct of point 3. If you have large number of glyphs or multiple fonts, the texture may become too big for your graphics API. **The typical result of failing to upload a texture is if every glyph or everything appears as empty black or white rectangle.** Mind the fact that some graphics drivers have texture size limitation. If you are building a PC application, mind the fact that your users may use hardware with lower limitations than yours.
Note that while OversampleH = 2 looks visibly very close to 3 in most situations, with OversampleH = 1 the quality drop will be noticeable. Read about oversampling [here](https://github.com/nothings/stb/blob/master/tests/oversample).
- Reduce glyphs ranges by calculating them from source localization data.
You can use the `ImFontGlyphRangesBuilder` for this purpose and rebuilding your atlas between frames when new characters are needed. This will be the biggest win!
- Set `io.Fonts.Flags |= ImFontAtlasFlags_NoPowerOfTwoHeight;` to disable rounding the texture height to the next power of two.
- Set `io.Fonts.TexDesiredWidth` to specify a texture width to reduce maximum texture height (see comment in `ImFontAtlas::Build()` function).
ImFont* font = io.Fonts->AddFontFromMemoryTTF(data, data_size, size_pixels, ...);
```
IMPORTANT: `AddFontFromMemoryTTF()` by default transfer ownership of the data buffer to the font atlas, which will attempt to free it on destruction.
This was to avoid an unnecessary copy, and is perhaps not a good API (a future version will redesign it).
If you want to keep ownership of the data and free it yourself, you need to clear the `FontDataOwnedByAtlas` field:
```cpp
ImFontConfig font_cfg;
font_cfg.FontDataOwnedByAtlas = false;
ImFont* font = io.Fonts->AddFontFromMemoryTTF(data, data_size, size_pixels, &font_cfg);
```
##### [Return to Index](#index)
---------------------------------------
## Loading Font Data Embedded In Source Code
- Compile and use [binary_to_compressed_c.cpp](https://github.com/ocornut/imgui/blob/master/misc/fonts/binary_to_compressed_c.cpp) to create a compressed C style array that you can embed in source code.
- See the documentation in [binary_to_compressed_c.cpp](https://github.com/ocornut/imgui/blob/master/misc/fonts/binary_to_compressed_c.cpp) for instructions on how to use the tool.
- You may find a precompiled version binary_to_compressed_c.exe for Windows inside the demo binaries package (see [README](https://github.com/ocornut/imgui/blob/master/docs/README.md)).
- The tool can optionally output Base85 encoding to reduce the size of _source code_ but the read-only arrays in the actual binary will be about 20% bigger.
Then load the font with:
```cpp
ImFont* font = io.Fonts->AddFontFromMemoryCompressedTTF(compressed_data, compressed_data_size, size_pixels, ...);
```
or
```cpp
ImFont* font = io.Fonts->AddFontFromMemoryCompressedBase85TTF(compressed_data_base85, size_pixels, ...);
Using an icon font (such as [FontAwesome](http://fontawesome.io) or [OpenFontIcons](https://github.com/traverseda/OpenFontIcons)) is an easy and practical way to use icons in your Dear ImGui application.
A common pattern is to merge the icon font within your main font, so you can embed icons directly from your strings without having to change fonts back and forth.
To refer to the icon UTF-8 codepoints from your C++ code, you may use those headers files created by Juliette Foucaut: https://github.com/juliettef/IconFontCppHeaders.
To make your icon look more monospace and facilitate alignment, you may want to set the ImFontConfig::GlyphMinAdvanceX value when loading an icon font.
- Dear ImGui uses imstb\_truetype.h to rasterize fonts (with optional oversampling). This technique and its implementation are not ideal for fonts rendered at small sizes, which may appear a little blurry or hard to read.
- There is an implementation of the ImFontAtlas builder using FreeType that you can use in the [misc/freetype/](https://github.com/ocornut/imgui/tree/master/misc/freetype) folder.
You can use the `ImFontGlyphRangesBuilder` helper to create glyph ranges based on text input. For example: for a game where your script is known, if you can feed your entire script to it and only build the characters the game needs.
As an alternative to rendering colorful glyphs using imgui_freetype with `ImGuiFreeTypeBuilderFlags_LoadColor`, you may allocate your own space in the texture atlas and write yourself into it. **(This is a BETA api, use if you are familiar with dear imgui and with your rendering backend)**
- You can use the `ImFontAtlas::AddCustomRect()` and `ImFontAtlas::AddCustomRectFontGlyph()` api to register rectangles that will be packed into the font atlas texture. Register them before building the atlas, then call Build()`.
- You can then use `ImFontAtlas::GetCustomRectByIndex(int)` to query the position/size of your rectangle within the texture, and blit/copy any graphics data of your choice into those rectangles.
- This API is beta because it is likely to change in order to support multi-dpi (multiple viewports on multiple monitors with varying DPI scale).
**Please note that many new C/C++ users have issues loading their files _because the filename they provide is wrong_ due to incorrect assumption of what is the current directory.**
Two things to watch for:
(1) In C/C++ and most programming languages if you want to use a backslash `\` within a string literal, you need to write it double backslash `\\`. At it happens, Windows uses backslashes as a path separator, so be mindful.
```cpp
io.Fonts->AddFontFromFileTTF("MyFiles\MyImage01.jpg", ...); // This is INCORRECT!!
io.Fonts->AddFontFromFileTTF("MyFiles\\MyImage01.jpg", ...); // This is CORRECT
```
In some situations, you may also use `/` path separator under Windows.
(2) Make sure your IDE/debugger settings starts your executable from the right working (current) directory. In Visual Studio you can change your working directory in project `Properties > General > Debugging > Working Directory`. People assume that their execution will start from the root folder of the project, where by default it often starts from the folder where object or executable files are stored.
```cpp
io.Fonts->AddFontFromFileTTF("MyImage01.jpg", ...); // Relative filename depends on your Working Directory when running your program!
io.Fonts->AddFontFromFileTTF("../MyImage01.jpg", ...); // Load from the parent folder of your Working Directory
You can also find this tool under `Metrics/Debuggers->Tools->UTF-8 Encoding viewer` if you want to paste from clipboard, but this won't validate the UTF-8 encoding done by your compiler.
(2) To encode in UTF-8:
There are also compiler-specific ways to enforce UTF-8 encoding by default:
- Visual Studio compiler: `/utf-8` command-line flag.
- Visual Studio compiler: `#pragma execution_character_set("utf-8")` inside your code.
- Since May 2023 we have changed the Visual Studio projects of all our examples to use `/utf-8` ([see commit](https://github.com/ocornut/imgui/commit/513af1efc9080857bbd10000d98f98f2a0c96803)).
Or, since C++11, you can use the `u8"my text"` syntax to encode literal strings as UTF-8. e.g.:
```cpp
ImGui::Text(u8"hello");
ImGui::Text(u8"こんにちは"); // this will always be encoded as UTF-8
ImGui::Text("こんにちは"); // the encoding of this is depending on compiler settings/flags and may be incorrect.
```
Since C++20, because the C++ committee hate its users, they decided to change the `u8""` syntax to not return `const char*` but a new type `const char_t*` which doesn't cast to `const char*`.
Because of type usage of `u8""` in C++20 is a little more tedious:
You can use the `Metrics/Debugger` window (available in `Demo>Tools`) to browse your fonts and understand what's going on if you have an issue. You can also reach it in `Demo->Tools->Style Editor->Fonts`. The same information are also available in the Style Editor under Fonts.
You can use the `UTF-8 Encoding viewer` in `Metrics/Debugger` to verify the content of your UTF-8 strings. From C/C++ code, you can call `ImGui::DebugTextEncoding("my string");` function to verify that your UTF-8 encoding is correct.
- Proggy Fonts, by Tristan Grimmer http://www.proggyfonts.net or http://upperbounds.net
- Sweet16, Sweet16 Mono, by Martin Sedlak (Latin + Supplemental + Extended A) https://github.com/kmar/Sweet16Font (also include an .inl file to use directly in dear imgui.)
