layer-video

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

Layer Video Generation

Layer 视频生成

Overview

概述

Video models take a prompt the way a director takes a shot list, not the way an illustrator takes a brief. A video prompt describes what changes over the clip. A prompt that only describes the scene yields a near-still image with drifting artefacts, which is the most common failure here.

Pick the use case first:

text_to_video

from nothing,

image_to_video

to animate an existing frame,

video_extend

to continue a clip,

video_editing

to alter one. The loop and spend gate are in the

layer

skill. Assembling finished clips into one deliverable is

layer-video-timeline

If a sibling skill named here is missing from your available skills, ask the user to install it (

npx skills add layerai/skills --skill <name>

); unattended, proceed from tool schemas and flag the gap.

视频模型接收提示词的方式更像导演接收镜头清单，而不是插画师接收创作简报。视频提示词描述的是片段中会发生什么变化。如果提示词只描述场景，生成结果往往会变成一张近乎静止、且带有漂移伪影的图像，这是这里最常见的故障。

先选择用例：从零开始生成使用

text_to_video

，为现有帧添加动画使用

image_to_video

，延续片段使用

video_extend

，修改片段使用

video_editing

。

loop

筛选和消耗确认机制位于

layer

skill 中。将完成的片段组装成单个交付物则使用

layer-video-timeline

。

如果此处提到的同级 skill 不在可用 skill 列表中，请让用户安装（

npx skills add layerai/skills --skill <name>

）；在无人值守的情况下，则依据工具 schema 继续执行，并标记这一缺失。

Anchor on an image whenever one exists

只要存在图像，就以图像为锚点

image_to_video

beats

text_to_video

almost every time a look has already been approved. The first frame fixes the character design, palette, and composition, so the model spends its capacity on motion instead of re-inventing the subject. Attach the still as

init_image

first_frame

depending on what the model accepts, and where a model takes both

first_frame

and

last_frame

, supplying both is the most reliable way to control where a shot ends up.

Generate the key frame with

layer-image

, approve it, then animate it.

当视觉风格已经获得认可时，几乎总是应该优先使用

image_to_video

，而不是

text_to_video

。第一帧能够确定角色设计、色板和构图，让模型将能力集中在运动上，而不是重新设计主体。根据模型支持的参数，将静态图像作为

init_image

或

first_frame

传入；当模型同时接受

first_frame

和

last_frame

时，同时提供二者是控制镜头最终位置最可靠的方法。

使用

layer-image

生成关键帧，确认无误后，再为其添加动画。

Writing the motion

编写运动描述

Name three things, in this order:

Subject motion: what moves, in what direction, at what speed. "The banner snaps taut in a gust, then settles" is a shot. "A banner" is a photograph.
Camera: static, slow push in, orbit, handheld drift, crane down. State "locked-off camera" when the camera must not move, because models add drift by default.
Rate: slow, deliberate, sudden, gradual over the full clip. Models otherwise pick a middling pace that reads as neither.

Model families disagree about conventions, so read

get_base_model

for the one you picked rather than reusing phrasing that worked on another family. Where a model exposes structured

video_effects

(entries with an uppercase

type

such as

ORBIT_360

DOLLY_IN

, plus an optional

weight

from 0 to 100), prefer that over describing the same move in prose: it is the parameter the model was tuned on.

按以下顺序说明三件事：

主体运动：什么在移动、朝哪个方向移动，以及移动速度。“横幅在阵风中猛然绷紧，随后逐渐静止”描述的是一段镜头；“一面横幅”描述的是一张照片。
镜头：固定机位（static）、缓慢推进（slow push in）、环绕（orbit）、手持漂移（handheld drift）或摇臂下降（crane down）。当镜头必须保持不动时，明确写出“locked-off camera”，因为模型默认会加入漂移。
节奏：缓慢、从容、突然，或贯穿整个片段逐渐变化。否则，模型通常会自行选择不慢不快的节奏，无法呈现出上述任何一种效果。

不同模型系列对提示词约定有所不同，因此应针对所选模型读取

get_base_model

，而不是直接复用其他模型系列中有效的表述。如果模型提供结构化的

video_effects

（条目包含大写的

type

，如

ORBIT_360

或

DOLLY_IN

，并可选提供 0 到 100 的

weight

），应优先使用该参数，而不是用自然语言描述相同的镜头运动，因为这是模型专门调优过的参数。

Duration, audio, and lip sync

时长、音频与唇形同步

Clip length is per-model and short, usually a handful of seconds. Do not plan a 30-second piece as one generation without confirming a model supports it.

Some models choose their own length and ignore

duration_seconds

entirely. When the clip has to land on an exact length, filter for

fixed_duration

, which excludes them. Without it a model may simply hand back a different length, and nothing in the response reads as an error.

generate_audio

is a capability, so filter for it when the clip needs native sound rather than assuming silence can be fixed later.

lipsync

is a separate capability for dialogue on a character.

filter.capabilities

is an object of booleans, not a list of names:

{generate_audio: true}

Video runs count against a tighter rate limit than images: 60 generations per minute per user, of which at most 30 may be video.

片段长度取决于具体模型，而且通常很短，往往只有几秒钟。在确认模型支持之前，不要将 30 秒的内容规划为一次生成。

有些模型会自行选择长度，并完全忽略

duration_seconds

。当片段必须达到精确时长时，应按

fixed_duration

筛选，这会排除上述模型。否则，模型可能只是返回不同时长的片段，而响应中不会有任何内容表明这是错误。

generate_audio

是一项能力，因此当片段需要原生声音时应按该能力筛选，而不要假设之后可以再修复静音问题。

lipsync

是为角色对白提供的独立能力。

filter.capabilities

是一个由布尔值组成的对象，而不是名称列表：

{generate_audio: true}

。

视频任务比图像任务受到更严格的速率限制：每位用户每分钟最多生成 60 次，其中最多可有 30 次为视频生成。

Multi-shot pieces

多镜头作品

An ad, trailer, or cutscene is several shots. Before assembling, check whether a model can render the whole thing in one generation, because one generation holds its own continuity and assembly does not. When assembly is genuinely needed:

Lock the look with one approved still.
Generate each shot as
```
image_to_video
```
from a frame consistent with that still.
Cut them together with
```
layer-video-timeline
```
.

Storyboard it first if there are more than three shots: see

layer-art-direction

广告、预告片或过场动画通常由多个镜头组成。在开始组装之前，应先检查模型是否能够在一次生成中渲染整个作品，因为单次生成本身能够保持连续性，而后期组装不能。确实需要组装时：

使用一张已确认的静态图像锁定视觉风格。
使用
```
image_to_video
```
，从与该静态图像保持一致的帧生成每个镜头。
使用
```
layer-video-timeline
```
将这些镜头剪辑到一起。

如果镜头超过三个，应先制作分镜脚本；参见

layer-art-direction

。

Loops

循环

Filter for the

loop

capability first. It selects models that compose a seam-free loop rather than merely describing one, which is a different thing from a model that will accept the word "looping" in a prompt and return a clip that jumps at the cut. Pair it with

fixed_duration

when the loop has to be a specific length.

Failing that, generate the anchor frame and use it as both

first_frame

and

last_frame

where the model accepts both, so the motion at least returns to where it started.

Either way, keep the motion cyclical by nature (a flag in wind, a torch flicker, a hovering idle) and avoid one-way motion such as a character walking out of frame, which cannot be made to loop by any amount of prompting.

首先按

loop

能力筛选。它会选择能够生成无接缝循环的模型，而不是仅接受提示词中的“looping”一词，却返回在剪辑点发生跳变的片段的模型。如果循环必须达到特定长度，应同时按

fixed_duration

筛选。

如果无法做到这一点，则生成锚点帧，并在模型接受

first_frame

和

last_frame

的情况下，将同一图像同时作为二者传入，使运动至少能够回到起点。

无论采用哪种方法，都应让运动在本质上具有周期性，例如风中飘动的旗帜、火炬的闪烁或悬停时的待机动作。避免使用角色走出画面等单向运动，因为无论怎样调整提示词，都无法将其制作成循环。

Worked example

实例

"Animate this approved character key frame into a 5-second idle loop."

The frame exists from a previous run, so it already has a
```
file_id
```
.

list_base_models

with

filter.use_case: "image_to_video"

and

filter.capabilities: {loop: true, fixed_duration: true}

, since this has to loop seamlessly and land on five seconds. Take the first result, then

get_base_model

to see whether it accepts

last_frame

and structured

video_effects

Prompt the motion only: "The character breathes slowly, cloak drifting in a light breeze, weight shifting subtly between feet. Locked-off camera. Gentle, continuous, no change in pose by the end of the clip."
Pass the same file as
```
first_frame
```
and
```
last_frame
```
so the loop closes.
Estimate. Video is priced well above images, so this will often clear 20 CUs and need an explicit yes.
Execute, poll at
```
poll_interval_seconds
```
, and check the seam by watching the end run into the start before calling it a loop.

“将这个已确认的角色关键帧制作成 5 秒的待机循环动画。”

该帧来自上一次运行，因此已经拥有
```
file_id
```
。
调用
```
list_base_models
```
，设置
```
filter.use_case: "image_to_video"
```
和
```
filter.capabilities: {loop: true, fixed_duration: true}
```
，因为该动画必须无缝循环并精确达到 5 秒。取第一个结果，然后调用
```
get_base_model
```
，确认模型是否接受
```
last_frame
```
以及结构化的
```
video_effects
```
。
只提示运动内容：“角色缓慢呼吸，斗篷在微风中飘动，重心在双脚之间细微转换。锁定机位（locked-off camera）。动作轻柔、连续，到片段结束时姿势保持不变。”
将同一个文件同时作为
```
first_frame
```
和
```
last_frame
```
传入，使循环闭合。
估算消耗。视频的计费远高于图像，因此消耗经常会超过 20 CUs，需要明确确认。
执行任务，按
```
poll_interval_seconds
```
轮询，并观看片段结尾衔接到开头的效果，确认接缝没有问题后，才能将其称为循环动画。

Common mistakes

常见错误

Prompting the scene instead of the motion, then reporting the model "barely moved".
Using
```
text_to_video
```
when an approved still already exists.
Leaving the camera unspecified, then fighting drift the model added by default.
Describing a camera move in prose when the model exposes
```
video_effects
```
.
Planning a long piece as one clip without checking a model's duration limit.
Assuming audio can be added later on a model that could have generated it natively.
Prompting for a loop instead of filtering for the
```
loop
```
capability.
Leaving
```
fixed_duration
```
off when the clip has to land on an exact length.
Calling a clip a loop without watching the seam.

提示词只描述场景而不描述运动，随后报告模型“几乎没有移动”。
已经存在获得确认的静态图像，却仍然使用
```
text_to_video
```
。
不指定镜头运动，随后又试图消除模型默认加入的漂移。
模型已经提供
```
video_effects
```
，却仍使用自然语言描述镜头运动。
未检查模型的时长限制，就将长内容规划为一个片段。
没有筛选原本能够生成原生音频的模型，却假设音频可以之后再添加。
仅仅在提示词中要求循环，而没有按
```
loop
```
能力筛选模型。
片段必须达到精确长度，却没有使用
```
fixed_duration
```
筛选。
没有观看片段接缝，就直接将其称为循环动画。