Local Batch Render

Point at a folder of images or clips with guidance and render many videos with a resume ledger and per-item logs. No cloud token needed.

What to say

Render this folder of photos as one video each with captions from the manifest.

How it works

Write input/batches/<name>/manifest.json (schemas/batch_render.schema.json)
npm run batch -- --manifest ... --validate-only
npm run batch -- --manifest ...

Example input (examples/batch_render.json)

{
  "batch_id": "example-batch",
  "output_dir": "output/batches/example-batch",
  "defaults": {
    "aspect": "9:16",
    "caption_style": "static",
    "burn_captions": true
  },
  "items": [
    {
      "id": "01-example",
      "scenes": [
        {
          "text": "Local batch rendering",
          "duration_seconds": 2,
          "motion": "none"
        }
      ]
    },
    {
      "id": "02-example",
      "scenes": [
        {
          "text": "Resume-safe vertical video",
          "duration_seconds": 2,
          "motion": "none"
        }
      ]
    }
  ]
}

At a glance

Best for

One image = one video at scale

You give

Images or clips and a manifest

You get

output/batches/<name>/

Time

Depends on the batch

Cost

Free

Good to know

uv run scripts/verify_batch_output.py <manifest> after the batch

Examples (real results, first 24 seconds)

No sample render of this one is checked in yet. Ask for it and the first result becomes the example.