Skip to content
Video

LTX-2.3 LoRA: ID card generation video tracking

1 min readupdated

A set of LoRA adapters for the LTX-2.3 model that turns standard video into a clip with ID cards. Every person in the frame is filled with a unique color that stays consistent throughout the entire clip. The model acts as a tracker without using extra motion detectors or complex object association algorithms. The solution is based on a research paper about generating stable identification colors.

Key features

  • Powered by LTX-2.3 22B dev with rank-32 weights (rank-32 LoRA).
  • Turns RGB video into clips with stable color tags for people.
  • Includes checkpoints trained on the DanceTrack and SportsMOT datasets.
  • Replaces classic detectors and motion models with a generative approach.

What you can do

  • Apply the adapters for multi-object tracking tasks in video streams.
  • Test coherent frame generation using scripts from the project repository.
  • Use the color palette to extract stable person identifiers.
  • Explore model behavior on sports or dance datasets.

Example

The adapters are used with the ltx_run_chain.py repository scripts to process video using a sliding window approach.

Read moreDetails
Benchmarks: 40.3 HOTA on the DanceTrack test, DetA 37.6
Hardware: about 0.4 frames per second on a single A100
Limitations: detection is the weak point, performs worst in crowds
License: LTX-2 Community License Agreement, non-commercial research only
Discuss Telegram 0

Will it run?

Runs in 4-bit

Full precision does not fit, a quantised version does: quality barely changes down to 8 bits, more noticeably at 4 bits.

  • BF16weights 43 GB · ready files53 GB
  • FP8 / INT8recommendedweights 22 GB28 GB
  • 4-bitweights 12 GB17 GB

Tip: go 8-bit.FP8 / INT8 is within about a percent of full precision in quality, takes half the memory and runs faster on RTX 40/50 and H100. Full precision only pays off for fine-tuning.

This is a LoRA: counted together with its base model LTX-2.3.

Estimate: weights plus working memory. Long videos, high resolutions and long contexts need more.

Run it on Massed Compute

The script downloads the model to the rented machine. How to run it is in the model card.

  1. 1. Rent a machine with a fitting GPU

    Open Massed Compute
  2. 2. Paste the script into the machine's terminal

    Connect over SSH (the Massed Compute panel shows the command) or open a terminal on the machine's desktop.

    HF_TOKEN / CIVITAI_TOKENThe script starts with token lines: paste yours between the quotes. The Hugging Face token is for gated models, the Civitai one for Civitai models.

    # Sonam5/Generator-as-the-tracker
    # ===== Tokens: paste yours between the quotes =====
    export HF_TOKEN=""        # huggingface.co/settings/tokens, for gated models
    # ==========================================
    mkdir -p ~/models
    curl -LsSf https://astral.sh/uv/install.sh | sh
    source "$HOME/.local/bin/env"
    uv tool install "huggingface_hub[cli]"
    hf download Lightricks/LTX-2.3 --local-dir ~/models/LTX-2.3
    hf download Sonam5/Generator-as-the-tracker --local-dir ~/models/Generator-as-the-tracker
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to LTX-2.3
  3. 3. Use it

    The model is in ~/models. How to run it is in the model card (link above).

Stop the machine in the panel when you are done: it bills by the minute while it runs.

How to speed it up

Picked for this model type. Choose your hardware above to see what fits you exactly.

  • SageAttention 3v2.2.0from October 28, 2025

    even faster on RTX 50 / Blackwell

    FP4 attention for Blackwell. Faster than SageAttention 2, but the authors suggest v2 when accuracy matters.

    How to enable: Build from the sageattention3_blackwell folder of the repo; in ComfyUI via KJNodes.

    Earlier generations (2)
    • SageAttention 2++ — Quantised (INT8/FP8) attention with no visible quality loss. The main speed-up for image and video generation.
    • SageAttention 1 — First version: INT8 for Q·K. Still works, but 2++ is faster and more accurate.
  • FlashAttention 44.0.0b40from November 18, 2026new release

    fastest exact attention on B200

    Kernels for Blackwell (beta). Exact attention without quantisation — fine for everything, LLMs included.

    How to enable: pip install flash-attn-4; vLLM and SGLang pick it up themselves.

    Earlier generations (3)
    • FlashAttention 3 — The Hopper (H100/H200) version with FP8.
    • FlashAttention 2 — The standard for RTX 30/40/50 and A100: exact attention without the memory cost of long contexts.
    • xFormers — The older way to speed up attention, works almost everywhere. Sage or Flash are faster where available.
  • cache-dit (DBCache)v1.5.6from November 20, 2026new release

    1.5–3× at nearly the same quality

    Reuses block outputs between diffusion steps. Supports most DiT models.

    How to enable: pip install cache-dit; one line cache_dit.enable_cache(pipe) for diffusers.

    Earlier generations (2)
    • MagCache — Skips steps by magnitude of change. More accurate than TeaCache, no calibration.
    • TeaCache — The first popular diffusion step cache. Works, newer methods are more accurate.
  • LightX2V step distill0.6.1from November 9, 2026new release

    5–10× fewer steps

    Distilled LoRAs and an engine: video in 4–8 steps instead of 30–50.

    How to enable: Load their step-distill LoRA, CFG 1.0, 4–8 steps.

  • torch.compile2.15.0from October 28, 2026new release

    10–30% for free

    Compiles the model for your GPU. Slower first run, faster afterwards.

    How to enable: model = torch.compile(model); in ComfyUI the TorchCompileModel node.

Model family

Requires the base model

Lightricks/LTX-2.3

This model is a LoRA / adapter. Download the base model first, then load this one on top.

New posts about versions of LTX-2.3 (quants, finetunes, LoRAs) will show up under the bell.

Quantised versions 12All on Hugging Face
Fine-tunes 12All on Hugging Face
LoRAs and adapters 12All on Hugging Face
Merges 1All on Hugging Face

About the model

Sonam5/Generator-as-the-tracker
Author
Sonam5
Category
Video
Released
Posted

On NeyroLearn

Hugging Face

00

Ask AI about this model

Answered by Gemini 4 · the chat is kept only in this tab

Comments