ltx-2.3LTX-2.3 is Lightricks' open-source text-to-video foundation model (released March 2026) — a diffusion-transformer (DiT) that generates high-fidelity video AND synchronized audio from one model, in a single pass, with no separate dubbing step. Its 4× larger gated-attention text connector sharply improves complex-prompt fidelity: multi-subject scenes, spatial relationships and style instructions are reproduced far more accurately. A remastered VAE delivers crisper detail, realistic textures and cleaner edges, while an upgraded vocoder produces clearer, better-synced audio. LTX-2.3 outputs native 1080p in both portrait (9:16) and landscape (16:9) — no cropping — at selectable 24 / 48 fps, and renders 5–20 second clips with full picture-and-sound delivered together, ready to ship. Through API Models you call it via one unified endpoint: send a prompt for text-to-video, or add one reference image for image-to-video. Billed per second by resolution (480p $0.02/s, 720p $0.04/s, 1080p $0.045/s) — cheaper than the official $0.06/s at every tier (up to ~67% lower at 480p, ~25% lower even at 1080p), pay-as-you-go with only successful requests charged.
4× larger gated-attention text connector — multi-subject scenes, spatial relations and style instructions reproduced accurately
Picture and sound generated together in one pass (upgraded vocoder) — ship without any post-dubbing
9:16 and 16:9 at true 1080p (no cropping), selectable 24 / 48 fps
Sharper fine detail, realistic textures and cleaner edges than LTX-2
Prompt only → text-to-video; add one image → image-to-video (auto-routed). 5-15s (T2V) / 5-20s (I2V)
Per second: 480p $0.02/s · 720p $0.04/s · 1080p $0.045/s — below the official $0.06/s at every tier (up to ~67% lower)
Generated video will appear here
Provide URLs and click Generate
LTX-2.3 Video 是由 API Models 提供的视频生成 API。LTX-2.3 is Lightricks' open-source text-to-video foundation model (released March 2026) — a diffusion-transformer (DiT) that generates high-fidelity video AND synchronized audio from one model, in a single pass, with no separate dubbing step. Its 4× larger gated-attention text connector sharply improves complex-prompt fidelity: multi-subject scenes, spatial relationships and style instructions are reproduced far more accurately. A remastered VAE delivers crisper detail, realistic textures and cleaner edges, while an upgraded vocoder produces clearer, better-synced audio. LTX-2.3 outputs native 1080p in both portrait (9:16) and landscape (16:9) — no cropping — at selectable 24 / 48 fps, and renders 5–20 second clips with full picture-and-sound delivered together, ready to ship. Through API Models you call it via one unified endpoint: send a prompt for text-to-video, or add one reference image for image-to-video. Billed per second by resolution (480p $0.02/s, 720p $0.04/s, 1080p $0.045/s) — cheaper than the official $0.06/s at every tier (up to ~67% lower at 480p, ~25% lower even at 1080p), pay-as-you-go with only successful requests charged. 通过 API Models 平台，您可以使用统一的 API 接口调用该模型，享受比官方更低的价格。当前定价：480p · per second: $0.02, 720p · per second: $0.04, 1080p · per second: $0.045, 720p · 5s: $0.2, 1080p · ~16s: $0.72。
快速生成品牌宣传视频，适用于广告投放和社交媒体推广。
为抖音、小红书等平台创建引人注目的短视频内容。
生成产品功能演示和使用教程视频，提升用户转化率。
制作课程讲解、知识科普等教育类视频，降低视频制作门槛。
LTX-2.3 Video 通过 API Models 平台调用，当前定价：480p · per second: $0.02, 720p · per second: $0.04, 1080p · per second: $0.045, 720p · 5s: $0.2, 1080p · ~16s: $0.72。相比官方价格，最高可节省 95% 的费用。
在 API Models 注册账号并获取 API Key，然后通过我们的统一 API 端点调用即可。我们提供详细的 API 文档和 cURL、Python、Node.js 代码示例。
API Models 通过聚合平台提供与官方相同的 LTX-2.3 Video 模型，价格低 60-95%。我们提供统一的 API 接口，无需分别注册各平台账号，一个 API Key 即可调用所有模型。
LTX-2.3 是 Lightricks 于 2026 年 3 月发布的开源文生视频基础模型(扩散 Transformer / DiT)。它最大的特点是用同一个模型在一次推理里同时生成视频与同步音频(无需后期配音)。相比上一代,它配备 4 倍更大的「门控注意力」文本连接器,显著提升对复杂提示词的理解——多主体、空间关系、风格指令的还原都更准确;重制的 VAE 带来更锐利的细节、更真实的纹理与更干净的边缘;升级的声码器让音频更清晰、更同步。
① 复杂提示词精准还原(4× 文本连接器);② 原生音画同步——画面与声音一次性生成,直接交付,无需另配音;③ 原生 1080p 横竖双画幅(16:9 与 9:16,竖屏无需裁剪);④ 24 / 48fps 多帧率可选;⑤ 5–20 秒整段音画一次输出;⑥ 开源、支持 LoRA 微调。在 API Models 上它还是「文生 + 图生」二合一:只填提示词即文生视频,加一张参考图即图生视频。
适合需要「画面 + 声音一步到位」的场景:竖屏短视频(抖音 / 小红书 / Reels / TikTok)、社媒广告与产品短片、带配乐/音效的预告片与片头、口播/解说类内容、游戏与动漫风格短片、批量化内容生产。原生竖屏 1080p + 同步音频让它特别适合移动端短视频;开源 + LoRA 适合需要定制风格/角色的团队。
便宜。官方 LTX-2.3 在 1080p 下为 $0.06/秒(约 16 秒 $1)。通过 API Models 按秒计费:480p $0.02/s、720p $0.04/s、1080p $0.045/s——每一档都比官方 $0.06/s 便宜(480p 省约 67%,即便 1080p 也省约 25%);此外你还获得:一个 key 调所有模型、国内可直连、免认证、只对成功请求计费、注册送 $0.3。
LTX-2.3 Video 支持：Text & Image to Video、Native Synchronized Audio、1080p Portrait & Landscape、24 / 48 fps、5-20s、Open Source (Lightricks)。完整参数与调用方式见 API Models 的 API 文档。
可以。API Models 提供可直接访问的统一 API,一个 API Key 即可调用 LTX-2.3 Video,无需分别注册官方账号、也无需自行处理官方接口的网络访问。
我们支持 Stripe（Visa、Mastercard 等国际信用卡）和支付宝付款。充值后积分即时到账。