Abiray/LTX2.3-10Eros-1.3-GGUF

LTX2.3-10Eros v1.3 GGUF

This repository contains GGUF quantizations of the LTX2.3-10Eros (v1.3) Image-to-Video model originally created by TenStrip.

These weights have been converted and compressed into GGUF format to allow high-quality, long-context video generation on constrained local hardware setups with lower VRAM requirements.

Available Quantizations

File Name	Size	Recommendation
10Eros_v1.3-Q3_K_M.gguf	11.1 GB	Ultra-low VRAM setups
10Eros_v1.3-Q4_K_S.gguf	13.2 GB	Great balance of speed and size
10Eros_v1.3-Q4_K_M.gguf	14.3 GB	Highly recommended for 16GB VRAM configurations
10Eros_v1.3-Q5_K_S.gguf	15.0 GB	Minimal quality degradation
10Eros_v1.3-Q5_K_M.gguf	16.1 GB	Excellent fidelity, recommended balance
10Eros_v1.3-Q6_K.gguf	17.8 GB	Near-lossless precision
10Eros_v1.3-Q8_0.gguf	22.8 GB	High fidelity, closest to native BF16 performance

---

Deployment & Usage

These files are designed to be run locally inside ComfyUI utilizing the GGUF integration.

File Placement

1. Place the downloaded .gguf files into your directory: ComfyUI/models/diffusion_models/ (or ComfyUI/models/unet/ depending on your version layout).
2. Ensure you have the ComfyUI-GGUF node suite installed via the ComfyUI Manager.
3. Use the Unet Loader (GGUF) node to load the model file.

⚠️ Critical Workflow Requirement: Version 1.3 is strictly designed to function alongside the DMD LoRA. It is highly recommended to use the official V5 DMD layout. You can acquire the json setup here: 10Eros_10SNodes_I2V_Basic_DMD_V5.json.

---

Model Description & Changelogs

This model relies on SulphurAI/Sulphur-2-base as its core framework. It utilizes a highly tailored, layer-scaled merge across variable steps instead of a standard uniform weight blend, which allows it to respect text prompts significantly better than loading standard LoRAs.

v1.3 Changelog

• Structural Remix: Restructured specifically to mimic the fluid behavior found in the original beta version.
• Artifact Suppression: Heavily addresses previous issues involving persistent text subtitles, "ghost anatomy," and unexpected transitions.
• Prompt Adherence: Explicit motion capabilities and raw compliance remain competitive. Prompting requires an exact, descriptive, and highly directive style identical to the native 22B Dev model.

v1.2 Changelog

• Leveraged tuned connector data to mitigate facial drifting issues and assist longer director/camera prompts.
• Infused sulphur experimental weights directly to hone explicit structural physics.

---

Prompt Optimization Engineering

LTX architectures feature minimal built-in self-reasoning when heavily conditioned; the first frame, subsequent motion curves, scene evolutions, and audio cues must be explicitly declared.

For maximum generation accuracy, use an uncensored LLM or Grok with the following foreword block to write your production scripts:

Generate a video scene script with a description based on the attached image for an LLM that has a tokenizer that uses interleaved attention to support long-context understanding that is fed into a multimodal video model. Strict specification, follow up to the word: No timestamps. No unnecessary embellishment. Output only plain English text and make it a copy box.

First, describe the image initial scene in concise natural language; subject(s), subject(s) appearance, subject(s) composition and pose, background, and context.

Next, formulate a naturally evolving scenario that would take place describing every moving body part, composition change, and manipulation from the uploaded initial frame that would be reflected in the video models post-latent evolution output. If the image is explicit or sexual in nature, use full anatomical terminology and spice it up slightly with visually representable erotic themes.

Center the prompt around this basic idea: [ concept ]

interweave this dialogue or sound concept into the scene with descriptions of voice tone followed by the lines delivered in quotations, in a temporal sequence between or during motions. Dialogue should be concise and non-rambling as it will take away from video quality: [ dialogue ]

Inside that prompt describe only notable audio and audio queues, both normal and explicit; background noise as well as foley and natural sounds. In a temporal sequence paired with coinciding motions. In the case of absent dialogue or soundscapes and only if background music is fitting; describe a fitting genre and melodic tone with matching mood.

Output only text following above instruction. Follow-up suggestions should be on the topic of expanding or changing motion or dialogue from the output text.

Credits

• Original Fine-tune & Merge: TenStrip

• Base Architecture: SulphurAI & Lightricks