wavespeed-batch

Generates batch image or video A/B variants on Wavespeed with per-variant cost tracking.

9|Updated May 8, 2026
One-click install
npx skills add https://github.com/simpletibr/simplicio-loop-marketing --skill wavespeed-batch-simpletibr
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: wavespeed-batch
Source: https://github.com/simpletibr/simplicio-loop-marketing/tree/main/.skills/wavespeed-batch
Command: npx skills add https://github.com/simpletibr/simplicio-loop-marketing --skill wavespeed-batch-simpletibr

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Testing multiple creative variants (hooks, palettes, subjects) one at a time is slow and expensive. This Skill expands a single base prompt into a full variant matrix, runs it in parallel on Wavespeed, and reports the cost of each variant so you can filter ideas cheaply before committing to a premium provider. ## Core Features & Use Cases - Variant Matrix Expansion: Injects each value from a variant axis (hook, palette, subject) into a base prompt placeholder and generates every combination. - Cost Control: Estimates per-variant cost from the model price table, enforces an optional cost cap, and logs usage to a JSONL file. - Deterministic Reproducibility: Derives per-variant seeds from a base seed so batches can be reproduced exactly. - Use Case: Test 5 different hook overlays on the same editorial portrait using flux-schnell, get back 5 artifacts with a total cost of $0.12, then send only the winning variant to Higgsfield for a cinematic render. ## Quick Start Run a batch test on Wavespeed with the base prompt 'Editorial portrait, soft light. Hook overlay: { hook }' across these 5 hook values in 1:1 using flux-schnell, and report the cost per variant.

Frequently Asked Questions about wavespeed-batch

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I run A/B tests on AI image generation prompts?

Define a base prompt with a placeholder like { hook }, provide a variant axis listing the values to test, and submit the batch to Wavespeed. Each variant gets its own artifact, seed, and recorded cost so you can compare results side by side.

How to test multiple hooks or palettes before an expensive video render?

Run the variant matrix on a cheap fast model like flux-schnell or sdxl-turbo first, pick the winning variant from the returned artifacts, then send only that winner to a premium provider such as Higgsfield. This avoids spending render budget on untested ideas.

Which Wavespeed models support batch image and video generation?

The skill supports flux-schnell and sdxl-turbo for images and wan-video for video. You select the model via the model parameter along with kind (image or video) and aspect ratio such as 1:1, 9:16, 4:5, or 16:9.

Can I cap the cost of a batch generation run?

Yes, set the cost_cap_usd parameter and the batch estimates per-variant cost from the model price table before submission. If the cap is hit mid-batch, submission stops, partial results are returned, and unattempted variants are marked as failed.

What happens when some variants in a batch fail?

Each failed variant is retried once, and if it fails again it is marked as failed in the results with successful and failed counts reported separately. If all variants fail, the underlying provider error is surfaced instead of a misleading success log.