Short clip in Timeline: when does it loop and when does it freeze?

In Timeline 1.0, the default render.pad_mode of auto loops a clip that is too short for its slot, but only when the shortfall is more than 0.25 seconds and the clip is at least 0.5 seconds long. Smaller gaps and very short clips freeze on the last frame instead. Set pad_mode to freeze or loop to override the rule.

The two outcomes look very different on screen, so it is worth knowing which you will get. A 0.2-second freeze is invisible. A 3-second freeze reads as a stall, and a looped action can read as a glitch.

Reproduced with the compiler

We compiled a 4-second first slot against sources of different lengths, with the other slot long enough, and each pad mode. The cases are in the table. Every case produced one soft warning with the pad mode in its context, and none failed. The loop case reported loop_count: 2 for a 3-second clip filling 4 seconds.

The warning codes are segment_source_short_padded for a freeze and segment_source_short_looped for a loop. Both say that the timeline position is unchanged, so the audio sync is safe either way.

Choosing a mode

Use loop for ambient footage where repetition is natural, such as water, smoke or a spinning product, and freeze for anything with a clear action or a face, where a repeated gesture looks wrong. Leave auto for mixed material.

Neither is a substitute for the right length. If the source came from a video model, regenerating at the needed duration is usually cleaner than looping; the Seedance post works through that trade.

Where the setting goes and what it cannot do

The field sits in the render block of the request: { "render": { "pad_mode": "loop" } }. The same block holds strategy and the chunk settings, and unknown values are rejected. Stills are never short, because a still is held for its slot.

Because the free plan call does not probe media, it cannot tell you which mode a clip would trigger. Probe the clips with video inspect first and compare each length with source_in + duration.