Duration, Frame Rate and File Size: The Technical Parameters of Video
What the technical specs of a generated clip mean, which of them affect perception, and how to prepare a file for a platform.
A generated clip arrives with a set of characteristics the user usually does not choose, yet they shape how it looks after publication. Understanding them avoids unpleasant surprises on the platform.
Frame rate
Determines smoothness of motion. Standard values run from twenty-four to thirty frames per second, which is enough for movement to read as continuous.
A higher rate produces the video look — hyper-smooth motion associated with home video rather than cinema. A lower one introduces visible stepping.
In practice: if a clip looks jerky, the cause is usually not the frame rate but the abruptness of the movement itself — the model took steps that were too large between states.
Duration
Affects three things at once: cost, artefact accumulation and reception on the platform.
- Cost scales proportionally: every second is frames that must be generated.
- Artefacts accumulate: by the end of a long clip the face has drifted noticeably.
- Platforms treat length differently: in short vertical formats, five seconds with clear motion beat ten sluggish ones.
Resolution
Video models usually work at lower resolution than image models: generating dozens of coherent frames costs more, and resolution is what gets sacrificed.
For most platforms that is sufficient — they recompress uploads anyway. The problem appears on a large screen, where the softness shows.
Upscaling video is possible but accumulates invented detail in every frame, and the result often looks worse than the original.
Codec and file size
The file a service returns is usually already compressed sensibly. There is no need to recompress before uploading: the platform will do it, and double compression degrades quality.
The exception is a file above the platform's limit. Then compress once, at the highest quality that fits.
Aspect ratio
Set by the source frame and unchanged during generation. If you need a vertical clip, crop the source image before animating — cropping finished video loses resolution and edge margin.
Sound
Most photo-to-video scenarios return a clip without an audio track. That is not an oversight: there is nothing to synchronise invented motion with.
Sound is added separately in editing. Note that if the person in the clip appears to speak, the mismatch between articulation and audio will show — better to avoid such scenarios or not to show the face large.
What to check before publishing
- Does the aspect ratio suit the platform without cropping the face?
- Does the duration fit the format?
- Does the clip loop without a visible seam, if it will play in a loop?
- Do the first second and a half contain visible movement?
- Is the file within the size limit without extra compression?
Storage
Keep the original file separate from platform-prepared versions. In a month you may need another format, and it is better to make it from the original than from an already recompressed copy.
Frequently asked
Should I recompress the file before uploading?
No. The platform will compress it anyway, and double compression degrades quality. Recompress only if the file exceeds the limit — then once, at maximum quality.
Why does the clip look jerky?
Usually not the frame rate but the abruptness of the movement: the model took steps that were too large. Specifying a slow, smooth tempo helps.
Does generated video have sound?
Usually not — there is nothing to synchronise invented motion with. Audio is added separately in editing.
- #видео
- #технические параметры
- #публикация