视频帧数计算器
为 AI 视频换算帧数、帧率和时长
这个帧数计算器为 AI 视频和动画流程，在帧数、帧率和片段时长之间做换算。填入任意两个值，第三个立刻算出来，用的是 Wan、AnimateDiff、ComfyUI 的视频节点和 Stable Video Diffusion 共同遵循的（帧数 - 1）/ 帧率这条惯例。
开源视频模型是按帧思考的，不是按秒。Wan 2.x 在 16 fps 下生成 81 帧，AnimateDiff 工作在 16 帧的上下文里，而 ComfyUI 的批量数就是原始的帧数。电影人则按秒和时间线帧率思考。这个计算器在两者之间做翻译，下面的补帧面板还会告诉你，RIFE 或 FILM 在走向 24 fps 时间线的路上，对你的帧数和帧率做了什么。
全部在你的浏览器里运行。你填的任何内容都不会被发送到服务器。
你的换算结果
5 秒
帧数：81
帧率：16 fps
时长：5 秒
直接用帧数除以帧率会得到 5.063 秒。可播放的时长跨的是「帧数 - 1」个间隔，这也是生成出来的片段，比简单除法算出来的结果短一个帧间隔的原因。
补帧
RIFE、FILM 这类补帧工具会在不改变时长的前提下把帧率翻倍：每个帧间隔里都插进新的中间帧。
输出帧率：32
输出帧数：161
时长：5 秒（不变）
时长 =（帧数 - 1）/ 帧率 | 输出帧率 = 帧率 x 倍数 | 输出帧数 =（帧数 - 1）x 倍数 + 1
目标：24 fps 时间线
从 16 fps 出发，精确的补帧倍数是 1.5 倍，能得到 24 fps 下的 121 帧，时长还是 5 秒。只支持整数倍的补帧工具，可以用 3 倍补到 48 fps，再按 2 比 1 丢帧，或者在剪辑软件里降到 24。
如果改成把原始帧直接按 24 fps 播放，片段会缩短到 3.333 秒，时长损失 33.3%。正是这个偏差，让直接扔进 24 fps 时间线的 AI 片段显得被加速了。
帧数参考表
常见片段时长对应的帧数，用的是「帧数 = 时长 x 帧率 + 1」。
| 片段时长 | 8 fps | 12 fps | 16 fps | 24 fps | 25 fps | 30 fps | 60 fps |
|---|---|---|---|---|---|---|---|
| 2 秒 | 17 | 25 | 33 | 49 | 51 | 61 | 121 |
| 4 秒 | 33 | 49 | 65 | 97 | 101 | 121 | 241 |
| 5 秒 | 41 | 61 | 81 | 121 | 126 | 151 | 301 |
| 8 秒 | 65 | 97 | 129 | 193 | 201 | 241 | 481 |
| 10 秒 | 81 | 121 | 161 | 241 | 251 | 301 | 601 |
工作原理
选好要求解哪个值、填上另外两个，计算器就套用「时长 =（帧数 - 1）/ 帧率」。那个减一很关键：一个片段可播放的长度，是第一帧和最后一帧时间戳之间的那段时间，所以 N 帧只跨 N - 1 个间隔。这就是为什么 Wan 在 16 fps 下用 81 帧给出一个 5 秒片段，也是为什么 AnimateDiff 的渲染结果，会比帧数除以帧率略短一点。补帧面板套用的是「输出帧率 = 帧率 x 倍数」和「输出帧数 =（帧数 - 1）x 倍数 + 1」，也就是 ComfyUI 里 RIFE 和 FILM 节点用的那套算术。
5 秒视频是多少帧？
这完全取决于帧率。五秒在 16 fps 下是 81 帧，在 24 fps 下是 121 帧，在 30 fps 下是 151 帧，每一个都含末尾那一帧。一个从帧数到秒的计算器，只有在你知道自己的模型以什么帧率生成时才给得出有用的答案：Wan 默认 16 fps，AnimateDiff 的流程常见在 8 到 16 fps，Stable Video Diffusion 则按设置在 6 到 14 fps 之间输出短爆发。
帧率和帧数的关系，同样支配着 AI 视频的预算算术。ComfyUI 里的时长是按帧设定的，而每多一帧都要花显存和生成时间，所以知道「16 fps 下的 5 秒片段需要 81 帧」，就等于知道该在时长框里填什么数字，以及在把任务排进队列之前它会花多少钱。
把 AI 的产出弄上剪辑时间线，是另外一半功课。剪辑师在 24、25 或 30 fps 下剪片，而一段 16 fps 的生成结果扔进 24 fps 的时间线，要么播得短了三分之一，要么因为混合帧而抖动。补帧是那个便宜的解法：1.5 倍的一遍能把 16 fps 精确带到 24，而 3 倍补到 48 fps，在按 2 比 1 丢帧之后能给出干净的 24 fps，还白送你一个 2 倍慢动作的选项。
适合谁
- 用 AI 做电影的人 在花掉 GPU 时间之前，把按秒衡量的镜头表换算成 Wan、AnimateDiff 或 ComfyUI 真正要的帧数。
- 剪辑师 算出那个能把生成素材落到 24 或 30 fps 时间线上、既不抖动也不会被意外变速的补帧倍数。
- 动画师 按 8 或 12 fps（一拍二）规划动画，并清楚每种片段时长对应的确切帧数预算。
Video Frames Calculator: the complete guide
It solves any one of frame count, fps, or duration from the other two using the (frames - 1) / fps convention, and models frame interpolation so you know the output fps and frame count before running RIFE or FILM.
For this workflow, the central problem is clear: open video models measure clips in frames while filmmakers plan in seconds, and the (frames - 1) / fps convention makes naive conversions come out wrong. Left unresolved, this creates downstream friction and slower decisions. The practical target is exact frame counts, frame rates, and durations that match what the model generates and what the timeline expects.
Limitation to keep in mind: It cannot know model-specific frame count constraints (like Wan's 4n+1 steps) beyond documenting them, and it does not predict encoder behavior after export; verify final timing in your NLE.
Advanced workflow: Advanced teams standardize one generation fps per project, log frame counts per shot in the shot list, and batch all interpolation at a single factor so every clip conforms to the timeline identically.
Step-by-Step Workflow
- Pick the value to solve for, then enter the two you know: model output is usually frames plus fps, a storyboard is usually seconds.
- Use the fps chips to match your model's native rate (16 for Wan, 8 to 16 for AnimateDiff) instead of guessing.
- Run the interpolation panel to plan the path from generation fps to timeline fps before generating anything.
- Copy the result into your shot list so frame budgets survive into the edit and the cost estimate.
Use Cases By Profile
- AI filmmaker: convert a 6 second shot into the exact frame count to type into a ComfyUI length field before queuing.
- Editor: pick the interpolation factor that lands 16 fps generations on a 24 fps timeline without retiming artifacts.
- Animator: budget frames for 12 fps animation passes and know the runtime cost of every added second.
Common Mistakes To Avoid
- Dividing frames by fps and wondering why every clip is one interval shorter than expected.
- Dropping 16 fps generations straight onto a 24 fps timeline and shipping the resulting judder or speed-up.
- Ignoring model frame-step constraints and requesting counts the model silently rounds.
Professional Best Practices
- Memorize the anchors: 81 frames at 16 fps is 5 seconds, 121 frames at 24 fps is 5 seconds.
- Interpolate before color and grain passes; interpolators behave better on clean frames.
- When a model constrains counts (4n+1 for Wan), pick the nearest valid count and recompute the true duration here.
Treat this tool output as a decision support layer, not a replacement for authorship. Great scripts are remembered for specific choices, emotional precision, and clarity of dramatic movement. Tools help by removing noise so your energy can go where it matters: character, conflict, escalation, and payoff. If you review outcomes after each pass and keep an explicit log of accepted changes, your workflow becomes faster and more predictable from draft to draft. That consistency is exactly what professional collaborators value: fewer surprises, clearer rationale, and a script that evolves with intent.
Extended FAQ
What formula does a video frames calculator use?
duration = (frames - 1) / fps. The minus one exists because N frames span N - 1 playable intervals. Frames from duration is the inverse: duration x fps + 1.
Why does Wan generate 81 frames?
Wan 2.x natively outputs 16 fps and targets 5 second clips: 5 x 16 + 1 = 81. The model also prefers 4n+1 frame counts, and 81 satisfies both.
What fps does AnimateDiff use?
AnimateDiff's motion modules were trained around 8 to 16 fps context windows, and most workflows render at 8, 12, or 16 fps, then interpolate up for delivery.
Is it better to generate at 24 fps or interpolate up to it?
Generating at 24 fps costs 50 percent more frames than 16 fps for the same duration. Most workflows generate at the model's native rate and interpolate, reserving native 24 fps for shots with fast motion where interpolation artifacts show.
How does frame count affect VRAM and generation time?
Roughly linearly for time and significantly for VRAM, since video models attend across frames. Halving fps at the same duration halves the frame budget, which is the main reason 16 fps generation plus interpolation is the default economy move.
What frame rate should the final export use?
Match your delivery target: 24 fps for cinematic work, 25 for PAL broadcast, 30 or 60 for social and screen capture. Decide the target first, then plan generation fps and interpolation factor backward from it.
常见问题
因为时长是（帧数 - 1）/ 帧率，不是帧数 / 帧率。片段可播放的长度跨的是帧与帧之间的间隔，而 N 帧只有 N - 1 个间隔。16 fps 下的 81 帧片段正好是 5.0 秒，不是 5.06 秒。
81 帧：5 x 16 = 80 个间隔，再加上末尾那一帧。这正是 Wan 的默认输出是 81 帧的原因，它在原生的 16 fps 下解码出来就是一个 5 秒片段。
如果你的工具支持小数倍，就按 1.5 倍补帧，时长完全不变。如果补帧工具只支持整数倍，就用 3 倍补到 48 fps 再隔一帧丢一帧，或者按 48 fps 导出让剪辑软件去归一。别只是变速：把 16 fps 的帧按 24 fps 播放，会砍掉三分之一的时长。
把目标秒数乘以你的输出帧率再加 1。16 fps 下的 4 秒片段是 65 帧；24 fps 下是 97 帧。有些节点还会把帧数限制在模型特定的步长上（Wan 用 81 这样的 4n+1 值），所以请就近取到合法的数值。
不会。补帧是往每个已有的帧间隔里填进新的中间帧，所以对 16 fps 下的 81 帧做一次 2 倍补帧，会得到 32 fps 下的 161 帧，时长仍然是 5 秒。只有当你之后变速、做速度斜坡或丢帧时，时长才会变。
