LTX-2-19B Text-to-Video 可直接根据文本生成连贯的电影感视频,重点强调时间稳定性、自然运动和概念清晰度。当场景具有明确的视觉创意、且运动用于强化意义而不是喧宾夺主时,效果尤其出色。
About this model
LTX-2-19B Text-to-Video 是一款先进的 AI 驱动模型,能够将描述性文本Промпт转换为电影级视频。该模型强调时间稳定性、自然运动和概念清晰度,确保生成的每个场景不仅符合视觉预期,还能在叙事上强化原本意图。它针对具有强烈视觉创意的场景进行了优化,让运动增强故事表达,而不会掩盖故事线。
该模型基于先进的生成技术,将视觉元素细致地编织成连贯的视频序列。它能够以克制的运动处理动态序列,营造真正的电影感,非常适合希望将文本概念变为现实的内容创作者、数字艺术家和营销人员。无论用于制作宣传素材、视觉叙事还是实验数字艺术,LTX-2-19B 都能在技术能力与创作灵活性之间取得出色平衡。
Cost analysis
| Provider | Cost | Notes |
|---|---|---|
| muapiapp | $0.6 per generation | muapiapp offers competitive pricing that is 20-50% more affordable while delivering high-quality and reliably cinematic video outputs. |
| Fal.ai | $0.8 per generation | muapiapp is 20-50% cheaper than Fal.ai, providing comparable or superior video generation quality at a better price point. |
| Replicate | $0.8 per generation | Replicate 的定价与 Fal.ai 相近,而 muapiapp 在不牺牲高级功能和视觉保真度的情况下,具有 20–50% 的成本优势。 |
muapiapp offers competitive pricing that is 20-50% more affordable while delivering high-quality and reliably cinematic video outputs.
muapiapp is 20-50% cheaper than Fal.ai, providing comparable or superior video generation quality at a better price point.
Replicate 的定价与 Fal.ai 相近,而 muapiapp 在不牺牲高级功能和视觉保真度的情况下,具有 20–50% 的成本优势。
** Competitor pricing is estimated based on similar model architectures and usage tiers.
Configuration schema
| Parameter | Type | Description | Default |
|---|---|---|---|
| Промпт | string | 描述Видео内容的文本Промпт。 | A busy open-plan office at night, completely empty except for hundreds of computer monitors left on. Each screen displays a different blinking text cursor on a blank document. As the video progresses, the cursors begin typing on their own—single words, then short sentences—filling the screens with fragments of emails, code, poems, and notes. Overhead lights flicker subtly, dust particles float through the air, and the camera slowly moves down the aisle between desks. The glow from the screens reflects off chairs and glass walls, creating a quiet, unsettling atmosphere. Cinematic realism, restrained motion, strong contrast between darkness and screen light, modern near-future tone. |
| Соотношение сторон | Enum (2 options) | The aspect ratio of the generated video | 16:9 |
| Длительность (секунды) | int | 生成Видео的Длительность(秒)。 | 5 |
| Разрешение | Enum (3 options) | 生成Видео的Разрешение。 | 720p |
描述Видео内容的文本Промпт。
A busy open-plan office at night, completely empty except for hundreds of computer monitors left on. Each screen displays a different blinking text cursor on a blank document. As the video progresses, the cursors begin typing on their own—single words, then short sentences—filling the screens with fragments of emails, code, poems, and notes. Overhead lights flicker subtly, dust particles float through the air, and the camera slowly moves down the aisle between desks. The glow from the screens reflects off chairs and glass walls, creating a quiet, unsettling atmosphere. Cinematic realism, restrained motion, strong contrast between darkness and screen light, modern near-future tone.The aspect ratio of the generated video
16:9生成Видео的Длительность(秒)。
5生成Видео的Разрешение。
720pDeveloper documentation
准备文本Промпт: 编写详细的视频场景描述。请务必加入必要的视觉线索和希望捕捉的氛围。
选择参数: 选择所需的Соотношение сторон(例如 16:9 或 9:16)、分辨率(例如 480p、720p 或 1080p)以及视频时长(5 到 20 秒)。
Входные параметры设置: 使用提供的 API 端点,并以所需的 JSON 格式Входные параметры参数。确保Входные параметры参数符合技术Входные параметры schema。
生成视频: 通过端点提交请求。处理完成后,模型会返回一个链接到生成视频的 URL。
查看并迭代: 观看生成的视频,并根据需要调整文本Промпт或参数,以获得理想的视觉Выходные данные。
Frequently asked
该模型最适合详细且生动的描述性Промпт,清楚概述场景的视觉元素和氛围。加入关键视觉线索,并指定特定的电影风格,可以获得最佳结果。
LTX-2-19B 支持 16:9 和 9:16 的Соотношение сторон,以及 480p、720p 和 1080p 的分辨率,能够满足各种显示需求。
通过 muapiapp 使用 LTX-2-19B 每次生成收费 $0.6,比提供类似能力但价格更高的 Fal.ai 和 Replicate 便宜 20-50%。
有。视频时长可以设置为 5 到 20 秒。如果未指定,默认时长为 5 秒。