xAI 视频模型 Imagine Video 1.5 新增图像与语音参考，支持文生视频与原生 1080p 生成
Imagine Video 1.5 with References Our best video model, now with text, image, and voice references — generating up to 1080p. Jul 31, 2026
- 1xAI 为视频模型 Imagine Video 1.5 增加图像与语音参考，同一角色可跨场景保持相同人脸和声音
- 2文生视频无需起始图像即可生成，原生 1080p 同时支持文生视频与图生视频两种路径
- 3多参考功能每次生成最多 7 张参考图，每张可锁定人脸、产品或位置，可换角色、换场景或只换动作
- 4图像与语音参考率先在美国向 SuperGrok Heavy 和 SuperGrok Plus 开放，几天内推广至全部档位
这次更新的实质是给视频生成补上可控性，角色、语音和场景都能被锁定复用，而不只是单条视频的画质提升。
- 影响：内容与营销团队可用参考图固定角色和产品形象，减少多镜头物料中人脸与声音不一致的返工；原生 1080p 可省去后期放大工序；参考功能先限美国部分高档订阅，接入节奏和成本会因档位而不同。
- 参考：可考虑选一个固定角色或产品的短视频场景，用一张角色图加语音样本测试一致性表现；值得先整理品牌参考素材清单，并对比订阅档位与 API 调用的成本。
- 依据：能力与开放范围为 xAI 官方公告口径，7 张参考上限、原生 1080p 与 API 支持均出自官方说明，暂无独立评测验证。
When we launched Imagine Video 1.5 last month, it was our best video model yet — better motion, better physics, and better audio. Today it goes further: image and voice references, video from a prompt alone, and native 1080p generation.
Image and voice references start today in the US for SuperGrok Heavy and SuperGrok Plus on grok.com/imagine and iOS, rolling out to all tiers over the next few days.
Text-to-Video and native 1080p
Describe the shot — no starting image needed. Text-to-video pairs our image generation with image-to-video. Native 1080p is now supported with text-to-video and image-to-video. Text-to-video and native 1080p are generally available on grok.com/imagine, iOS, and Android.
Voice consistency
Pass in a character image and a voice reference, and both hold — the same face and the same voice in every scene.
Multi-Reference
Each reference image locks one thing in place — a face, a product, a location. Keep a character and swap the scene, keep the scene and swap the character, or hold both and change only the action. Up to seven references per generation.
In the API
Image references, text-to-video, and native 1080p are live in the xAI API with our best video model, grok-imagine-video-1.5. Voice reference support is available on request.
python
来源：xAI：News · x.ai