Pandora
A model for generating realistic, high-quality images captured on a mobile phone.
It uses a two-stage training process: the first stage uses a low-quality dataset to help the model learn outlines, and the second stage uses a high-quality dataset to improve image quality and details.
Every image generated will surprise you.
I've enabled gate access.
You can submit an access request, which I will manually review.
If you would like to expedite the process, please pay me $100, and I will then grant you access.
For payment-related questions, you should email me.

- Prompt
- none
- Negative Prompt
- none

- Prompt
- none
- Negative Prompt
- none

- Prompt
- none
- Negative Prompt
- none

- Prompt
- none
- Negative Prompt
- none

- Prompt
- none
- Negative Prompt
- none

- Prompt
- none
- Negative Prompt
- none

- Prompt
- none
- Negative Prompt
- none

- Prompt
- none
- Negative Prompt
- none

- Prompt
- none
- Negative Prompt
- none

- Prompt
- none
- Negative Prompt
- none

- Prompt
- none
- Negative Prompt
- none

- Prompt
- none
- Negative Prompt
- none

- Prompt
- none
- Negative Prompt
- none
Use
Using a custom pipeline for inference yields the best performance
import torch
from diffusers import FluxPipeline
base_model = "black-forest-labs/FLUX.1-dev"
repo = "likewendy/pandora"
pipe = DiffusionPipeline.from_pretrained(
base_model,
custom_pipeline=repo,
trust_remote_code=True,
torch_dtype=torch.bfloat16
)
pipe.load_lora_weights(repo)
pipe.enable_model_cpu_offload()
img = pipe(
prompt="1 gril",
negative_prompt="low quality, deformed, watermark",
height=1024,
width=1024,
guidance_scale=3.2,
num_inference_steps=36,
generator=torch.Generator(device=pipe.device).manual_seed(42)
).images[0]
img.save("out.png")
Limitations
- This model is not intended to or capable of providing factual information.
- As a statistical model, this checkpoint may amplify existing social biases.
- The model may not generate output that matches the prompt.
- Tip following is significantly influenced by the prompt style.
Out-of-scope Use
This model and its derivatives may not be used
- in any manner that violates any applicable national, federal, state, local, or international law or regulation.
- For the purpose of exploiting, harming, or attempting to exploit or harm minors in any way; including but not limited to soliciting, creating, acquiring, or distributing child exploitation content.
- Creating or disseminating verifiably false information and/or content with the intent to harm others.
- Creating or disseminating personally identifiable information that could be used to harm individuals.
- Harassing, abusing, threatening, stalking, or bullying individuals or groups.
- Creating non-consensual nudity or illegal pornography.
- For fully automated decision-making, adversely affecting the legal rights of individuals or otherwise creating or modifying binding, enforceable obligations.
- Initiating or facilitating large-scale disinformation campaigns.
Responsible Artificial Intelligence
I am committed to the responsible development of generative AI technology.
- Before releasing Pandora, I evaluated and mitigated numerous risks in the Pandora model and service, including the generation of illegal content.
- I implemented a series of pre-release mitigations to help prevent third-party misuse.
- For example, the HuggingFace repository for the Pandora model includes inference filters for illegal or infringing content.
- Use of the model must comply with the terms of the Pandora license and be subject to filters or human review.
- We may randomly contact known deployers of the Pandora model to verify that filters or human review processes are in place.
- Downloads last month
- 35
Collection including likewendy/pandora
Collection
3 items
โข
Updated