Batch Generation & Image-to-Image

Picora’s batch generation lets you submit multiple prompt blocks in a single request, with optional reference image support for character / style consistency.

Quick start

curl -X POST https://api.picora.me/v1/aigc/generate-batch \
  -H "Authorization: Bearer $TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "jobs": [
      {
        "promptYaml": "model: flux-schnell\nprompt: girl in autumn forest\naspect_ratio: \"1:1\"",
        "count": 4
      },
      {
        "promptYaml": "model: flux-1.1-pro\nprompt: same girl in winter snow\naspect_ratio: \"1:1\"\nseed: 12345",
        "count": 4,
        "referenceImageUrl": "https://media.picora.me/abc.jpg"
      }
    ]
  }'

Plan limits

Plan	Max images per batch
Trial	4
Pro	20
Pro+	50

Only 1 concurrent batch per user to keep cost predictable; submit a second batch and you’ll get HTTP 409.

Character consistency best practice

1. Pick a seed for the first image and note it down
2. For follow-up batches, reuse the same seed + reference_image pointing to the first generated image
3. Adjust strength (0.6-0.8) to control how closely the output sticks to the reference

Failure handling

• Pre-deduct total cost (e.g. 4 × $0.01 = $0.04)
• Each generation failure auto-refunds 1 unit (-$0.01)
• Final status: completed (all ok) / partial (some failed) / failed (all failed)
• Cancel before dispatch: full refund

See API reference for /v1/aigc/generate-batch and /v1/aigc/batch-jobs/:id.