Verified: a month 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.
So this is the proper 1.0 release of the detail slider for Klein 9b. Takes your images from minimal and simplified on the negative end to more detailed and complex on the positive side.
Honestly, this version is just... better. Like, noticeably better than those proof-of-concept releases. The changes across the -10 to 10 range are way more pronounced now, and it plays nicer with different styles instead of randomly deciding to break things. File size went up a bit to make room for the improved training, but whatever, it actually works consistently now, which is the whole point.
The slider doesn't fight Klein 9b's weird way of handling detail anymore. It just does what you'd expect a detail slider to do, across pretty much the full range. Extremes might still get a little funky, but that's kind of par for the course.
This isn't experimental anymore, it's just... done. Well, done enough. Klein 9b's still evolving as a model, so if you find weird edge cases or figure out sweet spots for specific styles, yeah, let me know. But this should handle most stuff without needing babysitting.
Old description for v0.1 and v0.2:
An experimental detail slider for Klein 9b. Less than one megabyte in size. This adjusts the level of detail and intricacy in your images from minimal and simplified on the negative end to more detailed and complex on the positive side.
The slider works across a -10 to 10 range, though the effect is notably more subtle compared to detail sliders for other models. Starts to break down around 8 in strength.
Klein 9b handles detail differently than most image generators, so this is very much a work in progress as I figure out how to best nudge the model's output without fighting against its natural tendencies.
Current status: This is a proof of concept release. It does something, but it's not perfect and needs refinement. I'm releasing it now to gather feedback and see how it performs across different prompts and use cases in the wild.
Please share your results! If you use this slider, post your images and let me know what worked, what didn't, and what strength ranges gave you the best results. The more feedback I get, the better I can make future training runs. Klein 9b is still relatively new territory, so your real-world testing is genuinely valuable for improving this.