A controlled image-generation skill that translates Stable Diffusion workflows into Nano Banana Pro prompts and reference-image tasks.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install stable-diffusion-image-generation-editing-alternative
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install stable-diffusion-image-generation-editing-alternative using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is Stable Diffusion Image Generation & Editing Alternative?
This Openclaw Skills integration provides a practical alternative to Stable Diffusion, SDXL, WebUI, and ComfyUI for AI image generation and editing. It converts concepts such as checkpoints, LoRA style effects, ControlNet pose or structure constraints, negative prompts, denoising strength, dimensions, and batch requirements into natural-language instructions for the fixed public_model_nano_banana_pro model.
The skill does not run local WebUI or ComfyUI nodes and does not distribute third-party model weights. Instead, it extracts the visible intent of an original workflow, preserves authorized reference-image relationships, and submits tasks to the AI Hive API. Openclaw Skills users receive a migration-oriented workflow with clear reproducibility limits and validation guidance.
Stable Diffusion Image Generation & Editing Alternative Use Cases
- Migrate Stable Diffusion or SDXL image-generation workflows to a hosted Nano Banana Pro alternative.
- Replace negative prompts with explicit instructions describing unwanted anatomy, text, logos, artifacts, or backgrounds.
- Transfer pose, character identity, composition, or structural constraints using one or more authorized reference images.
- Translate ControlNet-style floor plans, line drawings, perspectives, and layout constraints into natural-language image prompts.
- Reinterpret LoRA-inspired color palettes, materials, lighting, and visual characteristics without claiming to reproduce a specific LoRA.
- Perform low-change product-photo editing while preserving the product outline, labels, logo, camera angle, and composition.
- Generate text-to-image or image-to-image outputs with explicit aspect-ratio requirements.
- Document migration decisions, reference-image order, model IDs, and task IDs for review and evaluation.
How Stable Diffusion Image Generation & Editing Alternative Works
- Identify the source workflow, including Stable Diffusion, SDXL, WebUI, ComfyUI, checkpoint, LoRA, ControlNet, negative prompt, seed, steps, CFG, sampler, denoise, dimensions, and batch settings.
- Record the original workflow and model licensing or authorization details before migrating it.
- Describe what each source node changes in the visible result rather than attempting a literal node-for-node translation.
- Convert technical parameters into semantic controls such as stability, style intensity, preservation level, and variation range; these values are not guaranteed to be architecturally equivalent.
- Prepare authorized reference images and specify their order and role, such as pose, identity, structure, product appearance, or visual style.
- Rewrite negative prompts as explicit prohibited elements and rewrite LoRA or ControlNet behavior as observable visual constraints.
- Submit the task through
imagegen.py, which calls the fixedpublic_model_nano_banana_promodel through the AI Hive API. - Save the prompt, reference-image sequence, model ID, and task ID, then retrieve task status and output details.
- Validate the result across multiple samples by checking control-image adherence, subject facts, copyright or licensing sources, and cross-sample stability.
- Treat the result as an approximate migration rather than a pixel-identical reproduction of local SD seeds or node graphs.
Stable Diffusion Image Generation & Editing Alternative Setup
Install the required Python HTTP client, initialize the skill, and retrieve tasks through the supplied script. Authentication requests connect only to https://ai-hive.iclip.cn/api.
pip3 install requests
python3 "$SKILL_PATH/scripts/imagegen.py" init --skill-name stable-diffusion-image-generation-editing-alternative
python3 "$SKILL_PATH/scripts/imagegen.py" task --task-id <taskId>
Generate an image by supplying a natural-language prompt and optional authorized reference images. The following patterns illustrate negative-prompt migration and aspect-ratio control:
python3 "$SKILL_PATH/scripts/imagegen.py" generate \
--prompt 'Create a professional 4:5 ceramic artist studio portrait with natural window light. Keep both hands visible and anatomically plausible. Exclude extra fingers, duplicated limbs, text, watermarks, brands, excessive retouching, plastic skin, and a cluttered background.' \
--param aspect_ratio=4:5
python3 "$SKILL_PATH/scripts/imagegen.py" generate \
--image ./authorized-pose.png ./approved-character.png \
--prompt 'Use the first image for standing pose and arm direction and the second for the authorized original character identity. Preserve the character facial structure, hairstyle, clothing, and age characteristics without copying the reference person, background, or brand.' \
--param aspect_ratio=4:5
The script does not launch WebUI or ComfyUI and does not execute checkpoints, LoRA files, or ControlNet plugins locally. Ensure reference images are authorized before submission.
Stable Diffusion Image Generation & Editing Alternative Data Schema & Taxonomy
The skill organizes migration data around the source workflow, semantic controls, references, and hosted task metadata.
| Data group | Contents | Purpose |
|---|---|---|
| Source workflow | Stable Diffusion, SDXL, WebUI, ComfyUI, checkpoint, LoRA, ControlNet, negative prompt | Documents the original search or migration context; names do not imply compatibility with hosted weights or plugins. |
| Semantic controls | Stability, style strength, preservation level, variation range, aspect ratio, dimensions, and batch requirements | Expresses visible intent without claiming exact equivalence to seed, steps, CFG, sampler, or denoise values. |
| Reference inputs | File paths and ordered image roles, such as pose, identity, structure, style sample, or product photo | Defines how authorized images influence the generated result. |
| Prompt instructions | Positive requirements and explicit exclusions | Converts node behavior and negative prompts into natural-language constraints. |
| Hosted metadata | Fixed model ID public_model_nano_banana_pro, task ID, prompt, and reference-image order |
Enables task tracking and partial reproducibility. |
| Validation record | Control-image adherence, subject facts, licensing or copyright source, and cross-sample stability | Supports migration acceptance beyond visual similarity from one output. |
The skill can preserve prompts, reference-image order, model IDs, and task IDs, but it cannot guarantee pixel-level parity with a local Stable Diffusion seed or node graph. No local checkpoint, LoRA, ControlNet, WebUI, or ComfyUI files are generated or executed.
Stable Diffusion Image Generation & Editing Alternative Advanced Features
- Fixed-model migration through
public_model_nano_banana_prorather than local model-weight execution. - Multi-reference prompting with explicit image ordering and role assignment.
- ControlNet-style pose, structure, floor-plan, perspective, and composition translation through natural-language constraints.
- Negative-prompt semantic conversion into clear prohibited elements.
- LoRA-inspired visual transfer focused on observable materials, palettes, geometry, and lighting without claiming exact LoRA reproduction.
- Low-denoise-style editing instructions for preserving product contours, labels, logos, colors, camera angle, and composition.
- Aspect-ratio control for portrait, landscape, square, and other requested output formats.
- Task-based execution and status retrieval using task IDs.
- Migration auditability through saved prompts, reference order, model IDs, and task IDs.
- Validation guidance that evaluates multiple samples, control adherence, subject accuracy, licensing, and stability rather than one-image resemblance.
- Clear architectural boundaries: no local WebUI, ComfyUI, checkpoint, LoRA, or ControlNet execution, and no chat, video, user, or balance features.
SKILL.md
Loading
Related Openclaw Skills
Commercial Product Background Replacement
Generate commercial-ready product background replacements while preserving the product’s exact identity, structure, branding and realistic scene integration.
Virtual Try-On Product Image Generation
Generate reviewable apparel-on-model product images while preserving authorized adult identity, body proportions, garment details, layering, and ecommerce composition.
Product Image Suite Generation & Editing
Generate and revise a fact-checked, channel-ready image suite for one SKU while preserving product identity, claims, and visual consistency.
ZoomInfo to Cargo
ZoomInfo to Cargo rebuilds a sample of an existing prospecting list in Cargo so teams can measure verified coverage, data quality, and migration cost before renewal.
Midjourney Image Generation & Editing Alternative
A Nano Banana Pro workflow that converts Midjourney prompts and image references into controllable, production-ready visuals without Discord or a Midjourney account.
Hyperliquid Market Data
A read-only market intelligence skill for querying live Hyperliquid prices, liquidity, funding, metadata, and historical datasets without API keys.
Featured*
