Job Ran Out of Memory

You will see this when a job needs more memory than its worker has:

> FFmpeg command failed: the job ran out of memory and FFmpeg was stopped before

> it finished (SIGKILL). Split the work into smaller jobs or use lower-resolution

> sources: every input is decoded at full size, so several 4K clips in one job

> need the most memory.

The job ends with status failed and no output file.

Cause

When a job combines inputs with filters (scaling, cropping, concatenating,

overlaying), FFmpeg decodes every input at its full resolution and holds frames

from all of them while it works. Memory grows with the number of inputs and with

their resolution. A handful of 1080p clips is fine. Eight or ten 4K clips in a

single concat can need more memory than one worker has.

Solutions

Split the job

Concatenate in batches, then join the results. For ten clips, run two jobs of

five, then a third job that concatenates the two outputs. Each job holds far

fewer frames at once.

Scale the inputs down first

If the final video is 1080p or smaller, there is no need to feed 4K into the

combining job. Run one job per clip that scales it down:

{
  "inputs": [{ "url": "https://example.com/clip-4k.mp4" }],
  "outputFormat": "mp4",
  "options": [{ "option": "-vf", "argument": "scale=-2:1080" }]
}

Then combine the smaller outputs.

Use lower-resolution sources

If you control the source files, export them at the resolution you need rather

than at 4K.

A related message

> FFmpeg crashed (SIGSEGV) before it finished

That is a crash inside FFmpeg rather than a memory limit. Try the job again. If

it happens again with the same inputs, email javid@ffmpeg-micro.com with the job

ID.