Sign In

Glassworker 🎨 Flux.1-D

8

77

7

5

Updated: Apr 21, 2025

styleanimeusman riazglassworker

Verified:

SafeTensor

Type

LoRA

Stats

77

7

38

Reviews

Published

Mar 1, 2025

Base Model

Flux.1 D

Training

Steps: 6,500

Usage Tips

Strength: 1

Hash

AutoV2
BCAC730FBE
seruva19's Avatar

seruva19

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.

Overview

"The Glassworker" is a Pakistan's first hand-drawn animated feature film. Directed by Usman Riaz and produced by Mano Animation Studios, the film is a coming-of-age story set against the backdrop of a country on the brink of war. It follows the lives of Vincent Oliver, a young glassblower, and Alliz Amano, a violinist, as they navigate love and personal growth amidst societal upheaval.

The animation style is notably influenced by Studio Ghibli, reflecting a blend of Western and South Asian cultural elements. The film premiered at the Annecy International Animation Film Festival on June 10, 2024, and was released in Pakistani theaters on July 26, 2024.

Usage

This LoRA does not require trigger words.

Images were generated with the following parameters (workflow embedded):

Model: flux1-dev (fp8_e4m3fn)
Text Encoder: t5pxxl_fp16/clip_l
Sampler: euler
Scheduler: normal
Steps: 24
Flux Guidance: 4
LoRA Strength: 1

Compatibility with other checkpoints and LoRAs was not tested.

About training

LoRA was created with the ai-toolkit (Windows 11, RTX 3090, RAM 64 Gb). Images used for training were extracted from official english trailer, total of 89 images. They were captioned with CogVLM2, caption prompt was "Describe this scene in a narrative manner."

Trained on flux-dev2pro-bf16-flux. Training parameters include:

rank/alpha: 32/32
optimizer: prodigy
optimizer_params.decouple: true
optimizer_params.use_bias_correction: false
optimizer_params.betas: [0.9, 0.99]
optimizer_params.weight_decay: 0.05
noise_offset: 0.1
lr: 1
lr_scheduler: "constant"