This tutorial explains why one prompt produces different results on different video models, and teaches a method to read a new model's preferred style instead of memorizing syntax that breaks on the next release. It builds on the show's earlier line: "The components are the language. The dialect is just the accent." The mechanism comes down to training captions. A text-to-video model learns from millions of video-and-caption pairs, and the writing style of those captions becomes its native language. The [HunyuanVideo paper](https://arxiv.org/pdf/2412.03603) describes using a large language model to rewrite user prompts "to conform to a standardized information architecture, akin to training captions," and the [Waver paper](https://arxiv.org/pdf/2508.15761) says rewriting exists "to align diverse user inputs as closely as possible with the captions used during model training." [PromptEnhancer](https://arxiv.org/pdf/2509.04545) shows this rewrite stage is standard pipeline design. Many platforms run that rewrite silently. Veo on [fal.ai](https://fal.ai/learn/devs/veo3-prompt-guide-master-google-video-generation) ships an Enhance Prompt toggle defaulted on; [Google Flow](https://www.mindstudio.ai/blog/how-to-use-google-flow-gemini-omni-video-editing) puts Gemini in front of Veo, and Google promotes [meta prompting](https://blog.google/products/gemini/meta-prompting-veo-gemini-tips/). Turn enhancement off to see a model's true dialect. Two rough families: cinematic prose ([Veo](https://deepmind.google/models/veo/prompt-guide/), [Veo 3.1](https://cloud.google.com/blog/products/ai-machine-learning/ultimate-prompting-guide-for-veo-3-1), [Sora 2](https://developers.openai.com/cookbook/examples/sora/sora2_prompting_guide)) versus terse motion-led phrasing ([Runway Gen-4](https://help.runwayml.com/hc/en-us/articles/39789879462419-Gen-4-Video-Prompting-Guide), [Luma Ray2](https://lumalabs.ai/learning-hub/best-practices), [Pika](https://pikartai.com/prompt/)). Runway drops negative prompt support; Veo keeps it. [Kling 3.0](https://blog.fal.ai/kling-3-0-prompting-guide/) flipped from terse to cinematic across versions, which is the whole argument for re-reading the guide. The [JSON debate](https://creativepossible.substack.com/p/should-you-prompt-in-json) lands on "organized thinking helps, the model doesn't read brackets." Bench your own shots and watch the [Artificial Analysis Text-to-Video Arena](https://artificialanalysis.ai/video/leaderboard/text-to-video), where mid-2026 leaders include HappyHorse-1.0, Dreamina Seedance 2.0, and Kling 3.0.