Text-to-Image
MLX
Safetensors
English
apple-silicon
flux
flux-3
image-editing
image-to-image
quantized
Instructions to use VisualInference/FLUX-3-Image-Klein-MLX-6bit with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- MLX
How to use VisualInference/FLUX-3-Image-Klein-MLX-6bit with MLX:
# Download the model from the Hub pip install huggingface_hub[hf_xet] hf download VisualInference/FLUX-3-Image-Klein-MLX-6bit --local-dir FLUX-3-Image-Klein-MLX-6bit
- Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- LM Studio
- Atomic Chat
Card: curated gallery, honest limitations
Browse files
README.md
CHANGED
|
@@ -35,15 +35,19 @@ flux3-mlx "a ceramicist's studio at golden hour, shelves of glazed bowls" --mode
|
|
| 35 |
|
| 36 |
## Gallery
|
| 37 |
|
| 38 |
-
|
| 39 |
|
| 40 |
| | |
|
| 41 |
|---|---|
|
| 42 |
-
| <br><sub>Edit: remove the car and its shadow</sub> | <br><sub>Three references -> one gallery wall</sub> |
|
| 43 |
+
| <br><sub>Edit: recolor the dress, keep the light</sub> | <br><sub>Bounding-box layout (boxes drawn on top)</sub> |
|
| 44 |
+
| <br><sub>Portra 400 portrait, overcast window light</sub> | <br><sub>Unretouched macro, Leica M6 on film</sub> |
|
| 45 |
+
| <br><sub>Vision3 250D road-movie still, 2.4:1</sub> | <br><sub>CineStill 800T, pushed, red halation</sub> |
|
| 46 |
+
| <br><sub>Upsampled from "night bus stop in the rain, CineStill 800T, one woman waiting under a red umbrella"</sub> | <br><sub>Two-color risograph poster: exact typography</sub> |
|
| 47 |
+
| <br><sub>1967 paperback cover, painted</sub> | <br><sub>Contact sheet with grease-pencil notes</sub> |
|
| 48 |
+
| <br><sub>4x5 still life, hard window light</sub> | <br><sub>Product macro: chrome and resin</sub> |
|
| 49 |
+
| <br><sub>Architectural interior, blue hour</sub> | <br><sub>65mm landscape, lone walker</sub> |
|
| 50 |
+
| <br><sub>Upsampled from 12 words: "a halved pomegranate on a plaster plinth, hard window light, Portra 400"</sub> | |
|
| 51 |
|
| 52 |
## Which repo?
|
| 53 |
|
|
|
|
| 129 |
- Text rendering can be inaccurate.
|
| 130 |
- Edits can change content outside the requested region.
|
| 131 |
- Prompt following depends on prompt style.
|
| 132 |
+
- What we saw in our own 27-prompt test (on the BF16 reference as much as here, so these are model traits, not
|
| 133 |
+
conversion artifacts): occasional hand and finger errors, and smooth, slightly waxy skin or a familiar "Flux face"
|
| 134 |
+
on some portraits. Detailed prompts (or `--upsample`) and a few seeds help; edits, multi-reference and layout
|
| 135 |
+
were reliably strong.
|
| 136 |
- 16 steps at about 7 s/step on an M4 Pro. 4K output is supported by the code but impractically slow on current Macs.
|
| 137 |
- No safety classifier is bundled. Under the license, Qualifying Users who use outputs commercially must apply
|
| 138 |
content filtering or review.
|