Generating images

You generate images by writing a prompt block — a fenced code block with prompt language, containing YAML. This keeps the prompt right next to the document it illustrates, versioned like any other content.

```prompt
model: flux-1.1-pro
aspect_ratio: 16:9
prompt: cyberpunk city, neon lights, rain
negative_prompt: blurry
output_count: 2
seed: 12345
```

Click Generate on the block (or call POST /v1/aigc/generate). Before anything is charged, the dashboard shows a cost confirmation with the exact credits and your current balance. The generated asset then appears inline.

Models

Five models are available, trading off speed, cost, and quality. The exact cost of each run is shown before you confirm — the relative cost below is only a guide.

Model	Provider	Speed	Relative cost	Max outputs	Aspect ratios
`flux-schnell`	Replicate	Fastest (sync)	Lowest	8	1:1, 4:3, 16:9, 9:16
`sdxl-lightning`	Replicate	Fast (sync)	Lowest	4	1:1, 4:3, 16:9
`sd3`	Replicate	Standard	Standard	4	1:1, 4:3, 16:9, 9:16
`flux-1.1-pro`	Replicate	High quality	Higher	4	1:1, 4:3, 16:9, 9:16, 21:9
`dall-e-3`	OpenAI	High quality	Higher	1	1:1, 16:9, 9:16

Prompt fields: model, prompt (required); aspect_ratio, negative_prompt, output_count, seed (optional); and for image-to-image, reference_image (a Picora image URL) + strength (0–1 blend).

Sync vs async

Fast models (flux-schnell, sdxl-lightning) return the finished image directly (status: "ready").
Slower models return status: "generating" with an assetId; the image appears when ready. The dashboard updates automatically; via API you poll GET /v1/aigc/assets/{id}/status.

Batch generation

To generate many images at once, use batch generation (POST /v1/aigc/generate-batch) — submit multiple prompt jobs, each with its own count. The whole batch is charged once up front; jobs run in parallel and you watch live progress. If some images fail, only the failed ones are automatically refunded.

Batch size is capped per plan: Trial 4 · Pro 20 · Pro+ 50 images.

Versions

Run the same prompt again and you get a new version of that asset rather than a duplicate — Picora tracks every generation of a prompt as a version history. One version is marked current; you can browse the thumbnail timeline and promote any version to current. Deleted assets can be restored.

Credits & cost

AIGC generation is paid from your Credit wallet:

Cost is calculated from the model and output count, and pre-charged — if your balance is too low, generation is rejected (insufficient_credit) and nothing is charged.
If a generation fails, the credits are refunded automatically.
Paid plans include a monthly credit grant; beyond that you top up as needed.

See Credit wallet for the full guide to grants, packages, and history.

AIGC overview — projects, templates, and the creator/output model.
AI image tools — post-process images (crop, background removal, OCR…).