Historiccolorsoonr Schnell

🚀 Historic Color Soon® (by A.C.T. Soon®)

Historic Color Soon® is a versatile visual and textographic generative model. It excels in generating high - quality images with a low step - count, producing realistic color film - like visuals, and creating images with a "historical" aesthetic and lifelike details.

🚀 Quick Start

Installation

To use Historic Color SOON® with the 🧨 diffusers python library, first install or upgrade diffusers:

pip install -U diffusers

Basic Usage

Then you can use FluxPipeline to run the model:

import torch
from diffusers import FluxPipeline

pipe = FluxPipeline.from_pretrained("Shakker-Labs/AWPortrait-FL", torch_dtype=torch.bfloat16)
pipe.to("cuda")
pipe.enable_model_cpu_offload() #save some VRAM by offloading the model to CPU. Remove this if you have enough GPU power

prompt = "HST style autochrome film photograph portrait of 1910 woman playing poker against a purple feathered dinosaur, the green-eyed woman has moderately blemished skin with visible lines and pores, she smiles, film grain, Kodachrome"
image = pipe(
    prompt,
    guidance_scale=1.2,
    num_inference_steps=4,
    max_sequence_length=256,
    generator=torch.Generator("cpu").manual_seed(0)
).images[0]
image.save("hstcolor1.png")

To learn more check out the diffusers documentation.

✨ Features

• Low Step - Count Quality: It can generate high - quality outputs with a low step - count (2 to 8 in most scenarios). A 4 - step inference at around 768x768 can routinely produce photorealistic outputs comparable to or better than Flux v.1 Dev.
• Realistic Color Film Aesthetics: It can create realistic images similar to color film analog photography, covering a wide range of iconic visual styles from Autochrome to Kodachrome and beyond.
• Historical Aesthetic and Lifelike Details: It produces visuals with a "historical" or "lived - in" aesthetic, good chromaticity and luminosity dynamics, and more lifelike textural/anatomical/skin details compared to other models at a similar step - count/resource expenditure.

📦 Installation

As mentioned above, to use the model with the diffusers library, you need to install or upgrade diffusers:

pip install -U diffusers

💻 Usage Examples

Basic Usage

import torch
from diffusers import FluxPipeline

pipe = FluxPipeline.from_pretrained("Shakker-Labs/AWPortrait-FL", torch_dtype=torch.bfloat16)
pipe.to("cuda")
pipe.enable_model_cpu_offload() #save some VRAM by offloading the model to CPU. Remove this if you have enough GPU power

prompt = "HST style autochrome film photograph portrait of 1910 woman playing poker against a purple feathered dinosaur, the green-eyed woman has moderately blemished skin with visible lines and pores, she smiles, film grain, Kodachrome"
image = pipe(
    prompt,
    guidance_scale=1.2,
    num_inference_steps=4,
    max_sequence_length=256,
    generator=torch.Generator("cpu").manual_seed(0)
).images[0]
image.save("hstcolor1.png")

📚 Documentation

Foundations/Methods/Tools

• Primary Base Model: [FLUX.1 - schnell](https://huggingface.co/black - forest - labs/FLUX.1 - schnell) by Black Forest Labs.
• Secondary Base Model: [Pixelwave Schnell V.1](https://huggingface.co/mikeyandfriends/PixelWave_FLUX.1 - schnell_01/) by HumbleMikey. The Safetensors & GGUF weights for it are found [here](https://huggingface.co/mikeyandfriends/PixelWave_FLUX.1 - schnell_01/tree/main) at HF, and/or via the model's CivitAI page [here](https://civitai.com/models/141592?modelVersionId = 778964).
• Finetuning: Mainly on around 500 color photographs taken during the 1900s and 1910s by Sergey Prokudin - Gorsky. You can explore his work at the [online archive](https://prokudin - gorsky.org/). The original glass - plate negatives are held at the Library of Congress in Washington, DC, USA.
• LoRA Fine - tuning: Full - rank LoRA fine - tuning of FLUX.1 - schnell on 130 photographs from the same dataset to create the Historic Color Schnell LoRA.
• LoRA Experiments and Merging: Experiments with training five more LoRAs, merging results using the 'flux_merge_lora.py' script from [Kohya - ss's sd - scripts git](https://github.com/kohya - ss/sd - scripts/tree/sd3), and converting LoRAs using the 'convert_flux_lora.py' script when necessary.
• Final Merging: Merging the resulting LoRA into the checkpoint finetuned earlier from Pixelwave Schnell v.1.
• Training Adapter: Ostris' training adapter for Schnell: [ostris/FLUX.1 - schnell - training - adapter](https://huggingface.co/ostris/FLUX.1 - schnell - training - adapter).

Trigger Words

You may use HST to bolster the vintage/autochrome effect. Though, in contrast with the Historic Color LoRA(s), the trigger does not appear to be quite as impactful towards the checkpoint.

Historical Note

Prokudin - Gorsky's color photography technique involved three photo - exposures using specialized color - spectrum filters (R.B.G.). His focus on developer and filter quality, experimentation, and glass - plate optimization led to high - fidelity and vivid color photography. The resulting prints have idiosyncratic color, light, and motion artifacts.

🔧 Technical Details

The model is based on two base models: [FLUX.1 - schnell](https://huggingface.co/black - forest - labs/FLUX.1 - schnell) and [Pixelwave Schnell V.1](https://huggingface.co/mikeyandfriends/PixelWave_FLUX.1 - schnell_01/). It uses finetuning on historical photographs, LoRA fine - tuning, and a series of merging operations. The training adapter [ostris/FLUX.1 - schnell - training - adapter](https://huggingface.co/ostris/FLUX.1 - schnell - training - adapter) is also used.

📄 License

The model is licensed under the Apache - 2.0 license.

Download Model

Weights for 'Historic Color SOON®' are available in Safetensors format in a separate repo. Download them in the Files & versions tab. The model is available in 🧨 Diffusers format within this repo.

Additional Information

Also, if you're into literature broadly and old modernist poetry specifically, check out my verse translations at [SILVER AGE POETS](https://www.SilverAgePoets.com/the - poets - and - their - stories)!