bria-ai/bria-skill/vgl
Claude Code and agent skill by bria-ai/bria-skill
- community
Checking that the SKILL.md file is still in the repository.
Maximum control over AI image generation — write structured VGL (Visual Generation Language) JSON that explicitly controls every visual attribute. Define exact object placement, lighting direction, camera angle, lens focal length, composition, color scheme, and artistic style as deterministic JSON instead of ambiguous natural language. Use this skill when you need reproducible image generation, precise control over scene composition, or want to convert a natural language image request into a structured JSON schema for Bria Fibo models. Triggers on requests for structured prompts, controllable generation, VGL JSON, deterministic image descriptions, or Bria/FIBO structured_prompt format.
Install
npx vskill@latest install bria-ai/bria-skill/vglSKILL.md
Bria VGL — Full Control Over Image Generation
Define every visual attribute as structured JSON instead of hoping natural language gets it right. VGL (Visual Generation Language) gives you explicit, deterministic control over objects, lighting, camera settings, composition, and style for Bria's Fibo models.
Related Skill: Use [bria-ai](../bria-ai/SKILL.md) to execute these VGL prompts via the Bria API. VGL defines the structured control format; bria-ai handles generation, editing, and background removal.
Core Concept
VGL replaces ambiguous natural language prompts with deterministic JSON that explicitly declares every visual attribute: objects, lighting, camera settings, composition, and style. This ensures reproducible, controllable image generation.
Operation Modes
| Mode | Input | Output | Use Case | |
|---|---|---|---|---|
| Generate | Text prompt | VGL JSON | Create new image from description | |
| Edit | Image + instruction | VGL JSON | Modify reference image | |
| Edit_with_Mask | Masked image + instruction | VGL JSON | Fill grey masked regions | |
| Caption | Image only | VGL JSON | Describe existing image | |
| Refine | Existing JSON + edit | Updated VGL JSON | Modify existing prompt |
JSON Schema
Output a single valid JSON object with these required keys:
1. short_description (String)
Concise summary of image content, max 200 words. Include key subjects, actions, setting, and mood.
2. objects (Array, max 5 items)
Each object requires:
Human subjects add:
Object clusters add:
Size guidance: If a person is the main subject, use "medium-to-large" or "large within frame".
3. background_setting (String)
Overall environment, setting, and background elements not in objects.
4. lighting (Object)
5. aesthetics (Object)
For people as main subject, specify shot type in composition: "medium shot", "close-up", "portrait composition".
6. photographic_characteristics (Object)
For people: Prefer "standard lens (35mm-50mm)" or "portrait lens (50mm-85mm)". Avoid wide-angle unless specified.
7. style_medium (String)
"photograph" | "oil painting" | "watercolor" | "3D render" | "digital illustration" | "pencil sketch"
Default to "photograph" unless explicitly requested otherwise.
8. artistic_style (String)
If not photograph, describe characteristics in max 3 words: "impressionistic, vibrant, textured"
For photographs, use "realistic" or similar.
9. context (String)
Describe the image type/purpose:
"High-fashion editorial photograph for magazine spread""Concept art for fantasy video game""Commercial product photography for e-commerce"
10. text_render (Array)
Default: empty array `[]`
Only populate if user explicitly provides exact text content:
Exception: Universal text integral to objects (e.g., "STOP" on stop sign).
11. edit_instruction (String)
Single imperative command describing the edit/generation.
Edit Instruction Formats
For Standard Edits (no mask)
Start with action verb, describe changes, never reference "original image":
| Category | Rewritten Instruction | |
|---|---|---|
| Style change | Turn the image into the cartoon style. | |
| Object attribute | Change the dog's color to black and white. | |
| Add element | Add a wide-brimmed felt hat to the subject. | |
| Remove object | Remove the book from the subject's hands. | |
| Replace object | Change the rose to a bright yellow sunflower. | |
| Lighting | Change the lighting from dark and moody to bright and vibrant. | |
| Composition | Change the perspective to a wider shot. | |
| Text change | Change the text "Happy Anniversary" to "Hello". | |
| Quality | Refine the image to obtain increased clarity and sharpness. |
For Masked Region Edits
Reference "masked regions" or "masked area" as target:
| Intent | Rewritten Instruction | |
|---|---|---|
| Object generation | Generate a white rose with a blue center in the masked region. | |
| Extension | Extend the image into the masked region to create a scene featuring... | |
| Background fill | Create the following background in the masked region: A vast ocean extending to horizon. | |
| Atmospheric fill | Fill the background masked area with a clear, bright blue sky with wispy clouds. | |
| Subject restoration | Restore the area in the mask with a young woman. | |
| Environment infill | Create inside the masked area: a greenhouse with rows of plants under glass ceiling. |
Fidelity Rules
Standard Edit Mode
Preserve ALL visual properties unless explicitly changed by instruction:
- Subject identity, pose, appearance
- Object existence, location, size, orientation
- Composition, camera angle, lens characteristics
- Style/medium
Only change what the edit strictly requires.
Masked Edit Mode
- Preserve all visible (non-masked) portions exactly
- Fill grey masked regions to blend seamlessly with unmasked areas
- Match existing style, lighting, and subject matter
- Never describe grey masks—describe content that fills them
Example Output
Common Pitfalls
- Don't invent text - Keep
text_renderempty unless user provides exact text - Don't over-describe - Max 5 objects, prioritize most important
- Match the mode - Use correct
edit_instructionformat for masked vs standard edits - Preserve fidelity - Only change what's explicitly requested
- Be specific - Use concrete values ("85mm portrait lens") not vague terms ("nice camera")
- Null for irrelevant - Human-specific fields should be
nullfor non-human objects
curl Example
References
- [Schema Reference](references/schema-reference.md) - Complete JSON schema with all parameter values
- [bria-ai](../bria-ai/SKILL.md) - API client and endpoint documentation for executing VGL prompts
Security
- Security scanPassed, 95 of 100Pattern checks on the skill's files.
- Model reviewPassed
Quality evaluation
- Task evaluationNot runNo evaluation yet. Run one in Skill Studio.
Source
- SKILL.md in the repositoryCheckingskills/vgl/SKILL.md
Popularity
- Repository stars
- 68
- Repository forks
- 6
- Repository npm downloads, 7 days
- 110
- 7-day trend score
- 34
Stars and forks are for the whole GitHub repository, not this skill alone. Downloads are for the repository's npm package. Refreshed 23 Mar 2026, 7 months ago.
Works with
Any agent vskill installs to: 56 targets in all. These 39 have a page on this site.
- Claude Code
- Cursor
- GitHub Copilot
- Windsurf
- Codex
- Gemini CLI
- Cline
- Amp
- Roo Code
- Goose
- Continue
- Kilo Code
- Junie
- Trae
- OpenHands
- Qwen Code
- AI Agent
- Tabnine Chat
- Sourcegraph Cody
- Amazon Q
- Firebase Genkit
- Sweep
- Devin
- Phind
- Mentat
- Claude Desktop
- Aider
- GPT Pilot
- Plandex
- bolt.new
- v0
- Replit Agent
- Anse
- TypingMind
- BetterChatGPT
- LibreChat
- Warp AI
- Pieces
- Fig AI
Embed badge
[](https://verified-skill.com/skills/bria-ai/bria-skill/vgl)