GLASNOST V.1: 80s USSR TV/Film

Style/Context Low Rank Adaptor (LoRA)
For Wan2.1 14B T2V & I2V Base Models
Stylers of Kinema Historical LoRAs
|||||||| By SilverAgePoets.com ||||||||

Prompt: [GLASNOST] style...

Prompt: [GLASNOST] style...

Prompt: [GLASNOST] style...

Prompt: [GLASNOST] style...

Prompt: [GLASNOST] style...

Prompt: [GLASNOST] style...

About this LoRA

This is a Rank 32/Alpha 64 LoRA for the Wan2.1 14b video generation model.

It was trained on hundreds of clips and frames from a variety of 1980s Perestroika-era Soviet films, tv shows, concerts, & music videos.

It can be used with diffusers or ComfyUI or DrawThings, etc...
This LoRA works well with both CausVid & Self-Forcing distillation quick inference adapters.
It also works fairly well in combos w/ other LoRAs.
Get creative with these!

Trigger words

You should use GLASNOST style vintage crisp analog footage from a 1980s soviet television movie, cinematic, video filmed in the USSR during the perestroika era, raw real life footage, vhs, etc, to ressurect one of these more recent gestalts of futures no-longer-past!

Using with Diffusers

pip install git+https://github.com/huggingface/diffusers.git

import torch
from diffusers.utils import export_to_video
from diffusers import AutoencoderKLWan, WanPipeline
from diffusers.schedulers.scheduling_unipc_multistep import UniPCMultistepScheduler

model_id = "wavespeed/Wan2.1-T2V-14B-Diffusers-fp16"
vae = AutoencoderKLWan.from_pretrained(model_id, subfolder="vae", torch_dtype=torch.float32)
pipe = WanPipeline.from_pretrained(model_id, vae=vae, torch_dtype=torch.bfloat16)
flow_shift = 3.0  # 5.0 for 720P, 3.0 for 480P
pipe.scheduler = UniPCMultistepScheduler.from_config(pipe.scheduler.config, flow_shift=flow_shift)
pipe.to("cuda")

pipe.load_lora_weights("AlekseyCalvin/Glasnost_v1_wan_14b_USSR80sTVstyle")

pipe.enable_model_cpu_offload() #for low-vram environments

prompt = "GLASNOST style"
negative_prompt = "overexposed, static, blurred, subtitles, images, static, worst, low, JPEG compression residue, incomplete, extra fingers, poorly drawn, poorly drawn, deformed, disfigured, misshapen, fused, still picture, backwards"

output = pipe(
    prompt=prompt,
    negative_prompt=negative_prompt,
    height=480,
    width=832,
    num_frames=81,
    guidance_scale=5.0,
).frames[0]
export_to_video(output, "output.mp4", fps=16)

Training details

Steps: 5000
Learning rate: 0.0002
LoRA rank: 32 dim, 64 alpha

Contribute your own examples

You can use the community tab to add videos that show off what you’ve made with this LoRA.