Menu
Use when generating images with Alibaba Cloud Bailian API, especially for Chinese text rendering or photorealistic images
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Based on SOC occupation classification
| name | imagenCN |
| description | Use when generating images with Alibaba Cloud Bailian API, especially for Chinese text rendering or photorealistic images |
| author | Agents365-ai |
| created | "2024-12-01T00:00:00.000Z" |
| updated | "2026-07-05T00:00:00.000Z" |
| homepage | https://github.com/Agents365-ai/imagenCN |
| metadata | {"openclaw":{"requires":{"bins":["python3"],"env":["DASHSCOPE_API_KEY"]},"primaryEnv":"DASHSCOPE_API_KEY","emoji":"🎨"}} |
Generate images using Alibaba Cloud Bailian API. Default endpoint is China region.
Supports five platforms across nine model families:
Cross-platform support: Windows, macOS, Linux
Automatically activate this skill when:
Users often give short, casual descriptions ("生成一只猫"). Before calling the API, present 3 refined prompt options with different style directions. Add, as appropriate:
Label the options clearly (e.g. A / B / C) with a one-line summary of each direction. Let the user pick one, combine elements from multiple, or request a new direction. Iterate until they confirm ("go", "generate", "ok", etc.), then proceed to generation.
Choose based on the request (see Model Selection Guide below). Default to
qwen-image-2.0-pro if unsure. Mention your choice to the user.
Native 2K for Qwen-Image 2.0, 1K/2K/4K for Wan2.7, or an aspect-ratio
preset (16:9, 1:1, etc.).
Run scripts/generate_image.py with the confirmed prompt and output path.
If the output path was implicit, save into the user's current working directory.
| Model | Description |
|---|---|
qwen-image-2.0-pro | Default. Latest flagship, native 2K, strongest typography and detail |
qwen-image-2.0-pro-2026-06-22 | Latest snapshot (Jun 2026): generation + editing fusion, better text rendering and prompt adherence |
qwen-image-2.0 | Standard 2.0 tier, native 2K |
qwen-image-max | Previous-gen flagship (Dec 2025) |
qwen-image-max-2025-12-30 | qwen-image-max snapshot: improved realism, fewer AI artifacts |
Editing models require an input image via --image (local path or URL). Omit --size to match the input image dimensions.
| Model | Description |
|---|---|
qwen-image-edit-max | Flagship editing model, strongest instruction following |
qwen-image-edit-max-2026-01-16 | Latest max snapshot (Jan 2026) |
qwen-image-edit-plus | Faster, lower-cost editing |
| Model | Description |
|---|---|
qwen-image-plus | Distilled accelerated version of qwen-image-max |
qwen-image-plus-2026-01-09 | qwen-image-plus snapshot (Jan 2026): faster high-quality generation |
qwen-image | Base model |
| Model | Description |
|---|---|
wan2.7-image-pro | Latest. Up to 4K output, unified architecture (T2I + edit + multi-image) |
wan2.7-image | Wan 2.7 standard, up to 2K |
wan2.6-t2i | Wan 2.6, flexible sizing |
wan2.5-t2i-preview | High quality, up to 768x2700 |
wan2.2-t2i-flash | Speed-optimized |
wan2.2-t2i-plus | Professional tier |
wanx2.1-t2i-turbo | Fast execution |
wanx2.1-t2i-plus | Professional tier |
wanx2.0-t2i-turbo | Earlier generation |
| Model | Description |
|---|---|
z-image-turbo | Fast, low-cost generation; bilingual (CN/EN) text rendering, high-fidelity portraits and product images. Pixel area 512x512 to 2048x2048 |
| Model | Description |
|---|---|
doubao-seedream-5-0-260128 | Ark default. Latest, up to 3K, PNG/JPEG output, best text rendering |
doubao-seedream-4-5-251128 | Seedream 4.5, up to 4K |
doubao-seedream-4-0-250828 | Seedream 4.0, up to 4K, budget-friendly |
| Model | Description |
|---|---|
hy-image-v3.0 | Hunyuan default. Flagship 3.0, strong composition awareness, handles complex Chinese prompts up to 8K chars |
| Model | Description |
|---|---|
cogview-4 | Zhipu default. Stable alias for latest CogView-4, native Chinese text rendering |
cogview-4-250304 | CogView-4 fixed snapshot (Mar 2025), reproducible results |
glm-image | GLM-Image flagship, up to 2048x2048, hybrid autoregressive/diffusion |
| Model | Description |
|---|---|
step-2x-large | StepFun default. High quality (0.1 RMB/image), up to 1024x1024 |
step-image-edit-2 | Fast & cheap (0.02 RMB/image), supports negative prompts, 8 inference steps |
# Default model (qwen-image-2.0-pro, native 2K output)
python ~/.claude/skills/imagenCN/scripts/generate_image.py "A cute cat" output.png
# Photorealistic with Wan model (Wan2.7 supports 4K)
python ~/.claude/skills/imagenCN/scripts/generate_image.py --model wan2.7-image-pro --size 4K "Realistic photo of mountains at sunset" photo.png
# Edit an existing image (requires --image; local path or URL)
python ~/.claude/skills/imagenCN/scripts/generate_image.py --model qwen-image-edit-max --image input.png "Change the background to a beach at sunset" edited.png
# Use ratio preset
python ~/.claude/skills/imagenCN/scripts/generate_image.py --size 16:9 "Wide landscape" landscape.png
# Use exact dimensions
python ~/.claude/skills/imagenCN/scripts/generate_image.py --size 1280*720 "Custom size" custom.png
Qwen-Image 2.0 (native 2K):
1:1 -> 2048x2048 (default)16:9 -> 2688x15369:16 -> 1536x26884:3 -> 2304x17283:4 -> 1728x23041K -> 1024x10242K -> 2048x2048Qwen-Image legacy:
1:1 -> 1328x132816:9 -> 1664x9289:16 -> 928x16644:3 -> 1472x11043:4 -> 1104x1472Z-Image (pixel area 512x512 to 2048x2048):
1:1 -> 1024x1024 (default)16:9 -> 1280x7209:16 -> 720x12802:3 -> 1024x15363:2 -> 1536x10241K -> 1024x1024Wan Series (Wan2.7 also accepts 1K/2K/4K):
1:1 -> 1024x10241:1-large -> 1280x128016:9 -> 1280x7209:16 -> 720x12804:3 -> 1200x9003:4 -> 900x12002:1 -> 1440x720Volcano Ark (Seedream):
1:1 -> 2048x204816:9 -> 2848x16009:16 -> 1600x28484:3 -> 2304x17283:4 -> 1728x23043:2 -> 2496x16642:3 -> 1664x24961K / 2K / 3K / 4K (model-dependent max resolution)Tencent Hunyuan (colon-separated format):
1:1 -> 1024:102416:9 -> 1920:10809:16 -> 1080:19204:3 -> 1600:12003:4 -> 1200:1600Zhipu (CogView-4 / GLM-Image):
1:1 -> 1024x1024 (default)16:9 -> 1344x7689:16 -> 768x13444:3 -> 1152x8643:4 -> 864x11522:1 -> 1440x7201:2 -> 720x1440StepFun (Step-2X):
1:1 -> 1024x1024 (default)1:1-small -> 512x51216:9 -> 1280x8009:16 -> 800x1280# With negative prompt
python ~/.claude/skills/imagenCN/scripts/generate_image.py --negative "blurry, low quality" "High quality portrait" portrait.png
# List all models
python ~/.claude/skills/imagenCN/scripts/generate_image.py --list-models
pip install dashscope requests
# Optional: for coloured output and styled tables
pip install rich
# Alibaba Cloud Bailian (DashScope)
export DASHSCOPE_API_KEY="your_api_key" # Required
export DASHSCOPE_MODEL="wan2.7-image-pro" # Optional default model
export DASHSCOPE_API_BASE="cn" # Optional: cn, sg, us
# ByteDance Volcano Ark
export ARK_API_KEY="your_api_key" # Required for Ark
export ARK_MODEL="doubao-seedream-5-0-260128" # Optional default model
# Tencent Hunyuan (TokenHub)
export HUNYUAN_API_KEY="your_api_key" # Required for Hunyuan
export HUNYUAN_MODEL="hy-image-v3.0" # Optional default model
# Zhipu / BigModel
export ZHIPUAI_API_KEY="your_api_key" # Required for Zhipu
export ZHIPUAI_MODEL="cogview-4" # Optional default model
# StepFun / 阶跃星辰
export STEP_API_KEY="your_api_key" # Required for StepFun
export STEP_MODEL="step-2x-large" # Optional default model
Get API Keys:
Create ~/.imagenCN.json for personal defaults, or .imagenCN.json in a project
directory for per-project overrides. API keys stay in environment variables for
security.
{
"platform": "ark",
"model": "doubao-seedream-5-0-260128",
"size": "2K"
}
All keys are optional. Priority (highest first):
--platform, --model, --size).imagenCN.json in current directory)~/.imagenCN.json)DASHSCOPE_MODEL, ARK_MODEL, HUNYUAN_MODEL)| Region | Alias | URL |
|---|---|---|
| China (default) | cn | https://dashscope.aliyuncs.com/api/v1 |
| Singapore | sg | https://dashscope-intl.aliyuncs.com/api/v1 |
| Virginia | us | https://dashscope-us.aliyuncs.com/api/v1 |
# Switch to Singapore endpoint
export DASHSCOPE_API_BASE="sg"
# Or use full URL
export DASHSCOPE_API_BASE="https://dashscope-intl.aliyuncs.com/api/v1"
| What you want | Model | Platform |
|---|---|---|
| Default / general (posters, text) | qwen-image-2.0-pro | DashScope |
| Photorealistic (portraits, landscapes) | wan2.7-image-pro | DashScope |
| Edit an image | qwen-image-edit-max | DashScope |
| Cheap & fast | z-image-turbo | DashScope |
| Photo + text combo | doubao-seedream-5-0-260128 | Volcano Ark |
| Complex Chinese composition | hy-image-v3.0 | Tencent Hunyuan |
| Chinese text in images | cogview-4 | Zhipu |
| Ultra-cheap volume gen | step-image-edit-2 | StepFun |
All other models are legacy/snapshot variants.
| Use Case | Recommended Model |
|---|---|
| General high-quality (default) | qwen-image-2.0-pro |
| Chinese text/calligraphy | qwen-image-2.0-pro |
| English text on images | qwen-image-2.0-pro |
| Posters with typography | qwen-image-2.0-pro |
| Photorealistic photos (4K) | wan2.7-image-pro |
| Photorealistic photos (2K) | wan2.7-image |
| Portrait photography | wan2.7-image-pro |
| Image editing (best quality) | qwen-image-edit-max |
| Image editing (fast, low-cost) | qwen-image-edit-plus |
| Fast, low-cost generation | z-image-turbo |
| High-fidelity portraits / product shots (fast) | z-image-turbo |
| Fast photorealistic (Wan) | wan2.2-t2i-flash |
| Lower-cost text rendering | qwen-image-plus |
| ByteDance best quality | doubao-seedream-5-0-260128 |
| Budget-friendly 4K (ByteDance) | doubao-seedream-4-0-250828 |
| Complex Chinese prompts (Tencent) | hy-image-v3.0 |
| Feature | DashScope | Ark | Hunyuan | Zhipu | StepFun |
|---|---|---|---|---|---|
| Best for | Text, variety | Photo+text | Complex CN | CN text in image | Ultra-cheap |
| Max res | 4K | 4K | 2K | 2K | 1K |
| SDK | dashscope | None | None | None | None |
| Price | Varies | ~0.22 | ~0.20 | ~0.06 | ~0.02 |
| Env var | DASHSCOPE_API_KEY | ARK_API_KEY | HUNYUAN_API_KEY | ZHIPUAI_API_KEY | STEP_API_KEY |
| Feature | ImagenCN (Bailian) | Imagen (Gemini) |
|---|---|---|
| Chinese text rendering | Excellent | Good |
| English text rendering | Excellent | Good |
| Photorealistic images | Excellent | Good |
| Speed | Medium | Fast |
| Model variety | 15+ models | 3 models |
| Max resolution | 4K (Wan2.7-Pro) | 2K |
# Default Ark model (Seedream 5.0)
ARK_API_KEY="xxx" python scripts/generate_image.py \
--platform ark \
"A vibrant close-up editorial portrait, Vogue magazine cover style" \
portrait.png
# With 4K output
ARK_API_KEY="xxx" python scripts/generate_image.py \
--platform ark --model doubao-seedream-4-5-251128 --size 4K \
"Breathtaking mountain sunset, golden hour, professional photography" \
landscape.png
# Default Hunyuan model (Image 3.0)
HUNYUAN_API_KEY="xxx" python scripts/generate_image.py \
--platform hunyuan \
"An astronaut riding a horse on the moon, cinematic lighting, 8K detail" \
scifi.png
# With prompt auto-enhance disabled
HUNYUAN_API_KEY="xxx" python scripts/generate_image.py \
--platform hunyuan --revise 0 \
"A cute orange cat napping in sunlight, oil painting style" \
cat.png
python ~/.claude/skills/imagenCN/scripts/generate_image.py \
"A beautiful Chinese New Year poster with red background, golden text, fireworks and firecrackers" \
new_year_poster.png
python ~/.claude/skills/imagenCN/scripts/generate_image.py \
--model wan2.7-image-pro \
--size 4K \
"Breathtaking sunset over mountain range, golden hour, professional photography" \
landscape.png
python ~/.claude/skills/imagenCN/scripts/generate_image.py \
--model wan2.7-image \
--size 2K \
"Professional product photography of a coffee cup on marble surface, studio lighting" \
product.png
Use when the user gives a topic and wants an automated topic-driven narrated explainer, podcast, or knowledge-summary video (Bilibili / YouTube / Xiaohongshu / Douyin / WeChat Channels), or asks to learn visual design patterns from a reference video/image. Trigger when the user mentions creating a knowledge video, narrated explainer, video podcast, or talking-head topic video from a topic — even if they don't say "video podcast" explicitly. Also trigger when the user wants to regenerate, re-render, rebuild, update, or iterate on a narrated video this skill already produced — e.g. they edited the script/prompt, changed the visuals, or swapped the background music and want the final video remade (reuse the existing videos/{name}/ directory, never start a new project). Do NOT trigger for generic video editing, trimming, format conversion, color grading, or non-narrative video tasks. Produces 4K video via research → script → TTS → Remotion → MP4 + BGM.
Multi-platform Chinese & multilingual TTS text-to-speech via Edge/Doubao/CosyVoice/Azure/Tencent/Baidu/MiniMax/Xunfei plus ElevenLabs/OpenAI/Google — 11 backends, word-level timestamps, [PAUSE:x] pause markers, pinyin pronunciation overrides
Use when looking up journal impact factors (JCR IF), checking a journal's impact factor by name, comparing IF across journals, or answering questions about "影响因子" / "impact factor" / "IF". Triggers on "impact factor", "journal IF", "影响因子", "JCR", "IF score", "journal rank", "which journal has higher IF", "what is the IF of". PROACTIVELY USE when user mentions journal prestige, publication venue quality, or manuscript submission target evaluation.
Use when looking up journal or magazine name abbreviations, converting between full names and ISO 4/MEDLINE abbreviations, processing BibTeX files for journal name standardization, or answering questions about 期刊缩写/杂志缩写. Triggers on "journal abbreviation", "abbreviate journal", "journal name", "期刊缩写", "杂志缩写", "ISO 4", "LTWA", "BibTeX journal". PROACTIVELY USE when user mentions citation formatting, reference list preparation, or manuscript submission to specific journals.
File new notes into the right folder and audit/reorganize folder structure in the user's Obsidian vault, using the `obsidian` CLI and a single source-of-truth map note (`00_Index/Folder_Map.md`) that lives inside the vault. Use this whenever a note needs to be placed, filed, sorted, or moved into the vault; whenever the user asks where a note "belongs" or "should go"; and whenever they want to clean up, reorganize, deduplicate, audit, or restructure vault folders (e.g. orphaned notes, dead-end notes, near-duplicate titles, overlapping folders). Trigger even when the user just says "add this to my vault", "put this somewhere sensible", or "tidy up the cellchat notes" without naming a folder. Requires the Obsidian desktop app to be running.
Use when designing, reviewing, or refactoring a CLI that must serve AI agents alongside humans, or when converting an API or SDK into an agent-usable CLI interface.