Download Latest Version
imference-desktop-go-windows-amd64-installer.exe (8.2 MB)
Add to Preferred Sources
| Name | Modified | Size | Downloads / Week |
|---|---|---|---|
| Parent folder | |||
| checksums.txt | 2026-10-01 | 438 Bytes | |
| imference-desktop-go-macos-universal.app.zip | 2026-10-01 | 13.0 MB | |
| imference-desktop-go-macos-universal.dmg | 2026-10-01 | 14.2 MB | |
| imference-desktop-go-windows-amd64-installer.exe | 2026-10-01 | 8.2 MB | |
| imference-desktop-go-windows-amd64.exe | 2026-10-01 | 15.3 MB | |
| README.md | 2026-10-01 | 1.8 kB | |
| v0.4.5 source code.tar.gz | 2026-10-01 | 2.8 MB | |
| v0.4.5 source code.zip | 2026-10-01 | 2.9 MB | |
| Totals: 8 Items | 56.3 MB | 12 | |
Imference Desktop v0.4.5
✨ New: LoRAs
Stack up to 4 LoRAs per generation to add a style, a character or extra detail.
- Curated catalog: browse LoRAs picked for each model family and download one in one click. Weights come pre-tuned, and the trigger words are shown when the LoRA needs them.
- Your own files: import any LoRA
.safetensors(e.g. from Civitai). The app detects which model family it was trained for and warns you before you use it on the wrong one. - Supported families: SDXL (incl. Pony and Illustrious), Z-Image, Krea 2 and Anima.
- Local and cloud: LoRAs work on your GPU and on imference.com cloud runs, at no extra cost.
- Picker dialog: the whole catalog is listed by family. LoRAs that don't fit the selected model are marked, and you can filter to show only the compatible ones.
✨ New: image-to-image for Anima
Anima models now accept a reference image, like SDXL, Z-Image or FLUX. Krea 2 is still text-to-image only.
🎛️ Interface
- Every composer card (Model, Local setup, Prompt, Format, Parameters, Reference image, LoRAs) now has the same header style, with a visible Add button.
- Help moved into tooltips that open on hover.
- One-line text fields no longer show a stray scrollbar.
⚙️ Engine
- Bundled inference engine updated to imference-engine v0.4.5: LoRAs on SDXL, Z-Image, Krea 2 (fp8 included) and Anima, plus Anima image-to-image.
- On first launch, the app upgrades an existing engine install automatically, so the first start takes a little longer.
Full changelog: https://github.com/Publikey/imference-desktop/compare/v0.4.3...v0.4.5
