Updated: Jan 30, 2026
conceptVerified: 5 days ago
SafeTensor
The FLUX.1 [dev] Model is licensed by Black Forest Labs. Inc. under the FLUX.1 [dev] Non-Commercial License. Copyright Black Forest Labs. Inc.
IN NO EVENT SHALL BLACK FOREST LABS, INC. BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH USE OF THIS MODEL.
Upscale small images
trained on pairs 256pix -> 1024pix
YOU must change standard workflow!
The input image (~256pix) must be connected directly to reference-condition (workflow inside PNG)
The use case is the upscale of small images with low resolution.
Always change the image size (your output resolution) by at least 2, preferably 4.
...
...
Flux2.D is quite ok but it is a large model.
Flux-klein-9b is in some conditions okay, this lora push and force upscale portrait images.
Main prompt:
preserve the overall impression of the image. Repair this portrait photo and refine.
oc you can add instructions eg "looking to the left", "change hair color to red" ...
Sometimes it helps to add a complete image description.
it is not that good for old analog photos, i will train next time.
9b-base-version:
Play with Lora-weight: 0.5 - 1
Play with CFG-Guider: 2 - 4
the distilled version (none base) 4-8 step with cfg 1, is also very impressive