bild: pure Go image processing from the CLI and as a library
Image processing algorithms in pure Go
At a glance
- What is it?
- bild is a collection of image processing algorithms written in Go, usable as a library or as a command line tool. It trades peak performance for a small dependency set and standard library image types.
- Who is it for?
- bild fits Go services that need resizing, blurring or channel work without pulling in a large imaging dependency, and it fits shell pipelines through the bild binary. It is the wrong choice if you need GPU acceleration, animated output or formats outside PNG, JPEG, BMP and WebP.
- Can I use it commercially?
- Yes. MIT is a permissive licence: you can use, modify and sell software built on it, as long as you keep its copyright and licence notices.
- Is it still maintained?
- Yes. The repository last received commits 69 days ago.
- What is it written in?
- Mainly Go, according to GitHub's language statistics.
Answers come from the project's GitHub data, last synced on November 15, 2026, and from our analysis. They are not legal advice.
Star history
- anthonynsimon/bild4,222 ★ · +9
| Repository | 2026-09-23 | 2026-11-15 |
|---|---|---|
| anthonynsimon/bild | 4.2k | 4.2k |
9 Stars in the last 53 days, measured from the daily GitHub snapshot. Compare with other projects
Editorial analysis
What bild solves, and who it is written for
Most Go projects that touch images end up binding to a C library or pulling a large dependency tree. bild takes the other route. The README describes it as "a collection of parallel image processing algorithms in pure Go" and states the aim plainly: "simplicity in use and development over absolute high performance". That sentence is the whole design brief. The project uses standard library packages wherever it can, and every operation returns a standard library image type, so results drop straight into code that already handles image.Image.
The audience follows from that. If you are writing a Go service that resizes uploads, applies a blur before thumbnailing, extracts an alpha channel, or runs an edge detection pass, bild covers the operation without cgo and without a separate runtime. The repository layout shows the scope: adjust/, blend/, blur/, channel/, clone/, convolution/, effect/, histogram/, imgio/, math/, noise/, paint/, parallel/, perlin/, segment/, and transform/. Each directory is a package, and the CLI in cmd/ and main.go exposes the same operations as subcommands.
It is not a Photoshop replacement and does not claim to be. What you get is a fixed menu of pixel operations with a consistent signature.
How the packages and the CLI divide the work
The library is the primary artefact. Each package exports functions that take an image and return a new one. The README's example chains three of them: effect.Invert, then transform.Resize with transform.Linear, then transform.Rotate with a nil options argument, and finally imgio.Save with imgio.PNGEncoder(). Nothing is mutated in place in that example; each call produces a new image.
Decoding and encoding sit in imgio. According to the README, imgio.Open decodes PNG, JPEG, BMP and WebP, and the available encoders are imgio.PNGEncoder(), imgio.JPEGEncoder(quality), imgio.BMPEncoder() and imgio.WEBPEncoder(options), where nil selects defaults. The CLI picks the encoder from the output file extension, so a .webp target writes WebP and a .jpg target writes JPEG.
Parallelism is the other structural choice. The README says most algorithms "make use of parallelism when available", and the parallel/ package exists for that purpose. Treat the parallelism as an implementation detail rather than a tuning surface.
The CLI itself is built on cobra, which appears as a direct dependency in go.mod alongside golang.org/x/image and github.com/HugoSmits86/nativewebp. The command tree in the README lists adjust, blend, blur, channel, effect, histogram, imgio, noise, segment and transform, each with its own subcommands and flags.
Installing bild and running a first transform
The README gives two routes. For the CLI you can compile from source with go get, or download a pre-compiled binary from the releases page. For the library, the install command is the module path with a trailing wildcard, and the README states that bild requires Go version 1.25 or greater.
go get github.com/anthonynsimon/bild/...After that, a build of the module produces the bild binary. The Makefile shows the intended local path: make build writes into dist/, and make install runs go install with the version linker flag.
make buildThe first real use is a resize. The README's own example takes input.png to output.png at 800x600 with the lanczos filter:
bild transform resize --width 800 --height 600 --filter lanczos input.png output.pngYou should end up with an output.png at the requested dimensions, written with the PNG encoder because of the extension. Two other transform subcommands are worth knowing on day one. Rotation with bound expansion uses --resize-bounds, and cropping takes a rectangle in the form 0x0+512x256:
bild transform crop --rect 0x0+512x256 input.png output.pngFor format conversion the README shows imgio encode, where the encoder follows the output extension:
bild imgio encode input.png output.webpIf you prefer the library, the README's basic example is the shortest path. It opens a JPEG, inverts, resizes to 800x800 with transform.Linear, rotates 45 degrees, and saves as PNG. Note that the example calls fmt.Println without showing the fmt import, so add it yourself.
Where bild stops being the right tool
The clearest limitation is stated by the project itself: simplicity over absolute high performance. A pure Go implementation of a Gaussian blur will not match a tuned C or SIMD implementation on large images, and the README makes no performance claim to the contrary. If your workload is batch processing thousands of large photographs, the per-image cost matters and a native library is likely the better fit.
Format support is narrow. The README lists four decodable formats and four encoders. There is no TIFF, no GIF, no AVIF, no HEIC, and no animated output of any kind. A pipeline that ingests phone photos in HEIC will need a conversion step before bild sees the file.
Parameter semantics are also under-documented in the README. Functions like effect.Median take a radius of 10.0 in the example, blur.Gaussian takes 3.0, and effect.Dilate takes 3. The README shows the call sites but does not explain the units or the valid ranges. For anything beyond the default, you are reading the package documentation on pkg.go.dev, not the README.
Finally, the CLI writes to a path you name.
bild against a native binding such as libvips
The obvious alternative for a Go service is a binding to a native imaging library, libvips being the common one. The difference is not a feature checklist, it is where the cost lands. It also requires the native library on every machine that builds or runs your code, which is exactly the thing bild avoids.
That makes the choice situational. A container image that already carries libvips, or a team comfortable with cgo and cross-compilation, gains throughput and format coverage. A small service, a CLI tool you distribute as a single static binary, or a build environment where cgo is disabled gains a great deal from bild's pure Go dependency set. The Makefile's release targets build linux and darwin binaries for amd64 and arm64 with -extldflags "-static", which is the kind of distribution story a cgo binding makes harder.
Within Go itself, golang.org/x/image is a dependency rather than a competitor, and it is already in bild's go.mod. If you only need one resize and nothing else, x/image alone may be enough. bild's value is the breadth of the effect, blend, channel and histogram packages on top of it.
Maintenance, licence and upgrade cost
The repository is not archived, and the last push was on 2026-09-07, the same day as the v0.17.1 release. The two prior releases, v0.17.0 and v0.16.1, landed on 2026-08-01 and 2026-07-09, so the release cadence over that window is roughly monthly. The project is still pre-1.0, which is the fact that should shape your upgrade planning: minor version bumps are where breaking changes are allowed to appear under semantic versioning conventions, and the CHANGELOG.md at the repository root is where the project records them.
The build requirement is a real upgrade cost of its own. go.mod declares go 1.25.0, and the README repeats the 1.25 floor. A team on an older toolchain cannot adopt bild without moving its Go version first, and that move affects the whole module, not just this dependency.
Licensing is MIT, per the badge and the LICENSE file at the root. MIT is permissive and imposes no copyleft obligation on your own code, but this is a description of the licence text, not legal advice. If you redistribute binaries built from bild, the usual MIT attribution requirement applies, and the release tarballs produced by the Makefile are the artefact to think about.
Editorial conclusion
bild fits Go services that need resizing, blurring or channel work without pulling in a large imaging dependency, and it fits shell pipelines through the bild binary. It is the wrong choice if you need GPU acceleration, animated output or formats outside PNG, JPEG, BMP and WebP. Before adopting, check the transform.Resize filter list in the package documentation against the interpolation quality your images need, and confirm that the encoder you plan to write with exists in imgio.
Frequently asked questions
What Go version does bild require?
The README states that bild requires Go version 1.25 or greater, and go.mod declares go 1.25.0. A project on an older toolchain has to upgrade before it can depend on bild.
Which image formats can bild read and write?
imgio.Open decodes PNG, JPEG, BMP and WebP. The encoders are imgio.PNGEncoder(), imgio.JPEGEncoder(quality), imgio.BMPEncoder() and imgio.WEBPEncoder(options), and the CLI chooses the encoder from the output file extension.
How do I install the bild command line tool?
The README gives two routes: compile from source with go get github.com/anthonynsimon/bild, or download a pre-compiled binary from the releases page. The Makefile also provides make build, which writes binaries into dist/.
How do I resize an image with bild on the command line?
The README example is bild transform resize --width 800 --height 600 --filter lanczos input.png output.png. The filter flag selects the interpolation method, and lanczos is the one shown in the documentation.
Official sources
Add this badge to your README
If you maintain this project, the badge below links readers to this analysis and shows its maintenance status from the daily GitHub snapshot. Paste the markdown into your README; add ?metric=license or ?metric=stars to the image URL for a different field.
[](https://hysenlabs.com/projects/anthonynsimon-bild)Star history chart for your README
[](https://hysenlabs.com/projects/anthonynsimon-bild)Was this useful?