autopostflow/gemma-4-e2b-ltx-2-5-prompt

VISIONPricing:Input $0.4 / Cached $0.08 / Output $0.8Concurrent Unit Cost:1Model Size:5.1BQuant:BF16Context Size:32kTool Calling:SupportedPublished:Sep 21, 2026License:gemmaArchitecture:Transformer Featherless Exclusive Cold

autopostflow/gemma-4-e2b-ltx-2-5-prompt is a 5.1 billion parameter fine-tuned Gemma 4 E2B IT model, specifically optimized for converting narrative stories into highly detailed LTX 2.5 cinematic video prompts. This text-only model, despite its multimodal base, excels at generating shot-by-shot video descriptions including camera, lighting, motion, and atmosphere. It is designed for users needing to transform textual narratives into structured prompts for LTX 2.5 video generators.

Loading preview...

Model Overview

autopostflow/gemma-4-e2b-ltx-2-5-prompt is a specialized 5.1 billion parameter model derived from google/gemma-4-E2B-it. It has been fine-tuned using a full bf16 process, focusing exclusively on text generation. While the base model is multimodal, this fine-tune freezes the vision/audio towers, making it a text-only model dedicated to a unique task.

Key Capabilities

  • Story-to-Prompt Conversion: Transforms narrative text into a single, highly detailed LTX 2.5 cinematic video prompt.
  • Cinematic Detail: Generates prompts that describe the most emotionally intense and visually compelling scene of a story, including shot-by-shot details like camera angles, lighting, motion, lens, and atmosphere.
  • Text-Only Processing: Efficiently processes text inputs to produce text outputs, despite its multimodal origins.

Training and Performance

The model was trained on 8958 examples, with loss computed only on the assistant's completion. It uses a maximum sequence length of 4096 tokens and was trained for 1.0 epoch with a learning rate of 0.0001. The training method ensures that the model learns to generate precise LTX 2.5 prompts from diverse story inputs.

Intended Use

This model is specifically intended for generating LTX 2.5 video prompts from narrative text. Its output can be directly fed into an LTX 2.5 video generator to create cinematic video content based on a story. Users should ensure the input prompt layout, particularly the ### CONTEXT separator, matches the training format for optimal results.

Finetunes
1 models