GroupCombo Tutorial: From One Request to a Reusable Content Template
Camika
GroupCombo is the “combined content template workbench” in CamikaLabs. It is designed to turn a complex creative request into multiple content groups, such as body text, titles, cover images, supporting images, video clips, tags, and more, then execute them together and combine the results into a complete output.
Unlike ordinary one-time generation, GroupCombo focuses on creating a reusable template first, then using that template repeatedly.
What It Is Good For
GroupCombo is suitable for:
- Xiaohongshu-style image-text posts
- WeChat official account articles
- Product marketing pages
- Multi-image social media content
- Tutorial articles with images
- Event promotion materials
- Brand content packs
- Combined text + image + video content
Example:
Create a spring skincare tutorial for Xiaohongshu, including a title, body text, cover image, 3 step-by-step images, and hashtags.
GroupCombo will split this into multiple content groups instead of generating only one paragraph of text.
Basic Workflow
1. Create a GroupCombo Template
In GroupCombo, start by entering your full content request. The clearer the request, the more stable the generated template will be.
A good request should include:
Content topic
Target platform
Required outputs
Style and tone
Target audience
Whether images or videos are needed
Example:
Create a Xiaohongshu-style city weekend travel guide template for young users aged 20-30.
It should include an eye-catching title, guide body text, cover image, 3 attraction images, closing engagement copy, and hashtags.
The tone should be relaxed, practical, and shareable.
After submission, the system starts the template creation process.
2. The System Generates a Blueprint
GroupCombo first uses an LLM to analyze the request and generate a blueprint. The blueprint defines which content groups the template contains.
It may produce a structure like:
cover_image Cover image
main_title Title
travel_guide Guide body text
spot_images Attraction images
closing_text Closing engagement copy
hashtags Hashtags
Each group has a type:
text
image
video
If one group needs to reference previous content, the system can create a dependency. For example, an image group may reference the guide text so the images better match the article.
3. The System Generates PE for Each Group
After the blueprint is confirmed, GroupCombo generates prompt-engineering data for every group, including:
system_prompt
user_prompt_template
variables
title
This allows each content group to know exactly what it should generate.
For example, the body group writes the guide, the cover image group creates the visual cover, and the hashtag group generates platform tags. They are generated separately instead of being mixed into one output.
4. Fill In Scene Variables
Once the template is created, the system extracts editable scene variables and turns them into a form.
You may see fields like:
City name
Trip duration
Target audience
Budget range
Style keywords
Reference image
Product name
Campaign theme
Each time you run the template, you only need to change these variables to create new content.
For example, the same travel guide template can be executed with:
City name: Hangzhou
Trip duration: 2 days
Style keywords: relaxing, healing, photo-friendly
Then next time:
City name: Chengdu
Trip duration: 3 days
Style keywords: food, slow-paced, youthful
This is what makes GroupCombo more valuable than one-time generation.
5. Execute the Template
After filling in the variables, click execute. The system creates an execution and starts running all groups.
Execution follows dependency rules:
- Groups without dependencies can run in parallel
- Groups depending on earlier content wait for those groups to finish
- Image groups can reference text content
- Image-to-image groups can reference previously generated images
- Upload groups directly use user-uploaded assets without calling a generation model
The results are saved per group and finally combined into a complete result page.
6. Review and Adjust the Results
After generation, you can review the result of each group.
Common actions include:
View body text
View images
Set cover
Edit prompt
Retry a single group
Re-execute the template
Delete history
Export results
If only one image is unsatisfactory, you do not need to rerun the entire task. You can retry only that specific group.
Complete Example
Goal: Create a Xiaohongshu skincare tutorial template.
Input request:
Create a Xiaohongshu skincare tutorial template for women aged 25-35.
Each execution should let the user fill in skin type, season, core product, and skincare goal.
Outputs should include: title, cover image, tutorial body text, 3 step-by-step images, tips, and hashtags.
The tone should be professional but friendly, suitable for saving and sharing.
GroupCombo may generate:
main_title text Generate a Xiaohongshu title
cover_image image Generate a tutorial cover image
intro_text text Generate opening recommendation copy
routine_steps text Generate skincare routine steps
step_images image Generate 3 step-by-step images
tips_text text Generate usage tips
hashtags text Generate hashtags
Scene variables may include:
skin_type Skin type
season Season
product_name Core product
skin_goal Skincare goal
tone Tone
ref_image Reference image
For execution, fill in:
Skin type: Combination oily skin
Season: Summer
Core product: Lightweight hydrating serum
Skincare goal: Oil control, hydration, non-comedogenic care
Tone: Relaxed, trustworthy, like a friend’s recommendation
The final output will be a complete Xiaohongshu-style image-text content package, not just an isolated paragraph or image.
Tips for Writing a Good Request
Good Request
Create a Xiaohongshu cafe review template.
The target users are young people aged 20-30 who enjoy photography and weekend outings.
Each execution should allow users to fill in city, cafe name, style, recommended drink, and average price.
Outputs should include title, cover image, body text, 3 ambience images, recommendation reasons, and hashtags.
The tone should be relaxed, lifestyle-oriented, and save-worthy.
Weak Request
Help me make Xiaohongshu content.
This is too vague. The system does not know the topic, platform, number of images, tone, or audience.
Usage Suggestions
Use GroupCombo for scenarios where you need to repeatedly produce the same kind of content. For example, weekly product notes, review posts, tutorials, marketing assets, and branded content.
If you only need a temporary paragraph of copy, a single generation tool is enough.
If you need multiple related outputs, GroupCombo is the better choice.
FAQ
What Is the Difference Between GroupCombo and Combo?
Combo is more like a one-time combined task. GroupCombo is a template workbench: create the template first, then fill in variables and execute it repeatedly. In the current project, GroupCombo is the main version, while Combo remains in the code but is hidden from Desktop.
Why Generate a Blueprint First?
The blueprint plans the content structure. Without it, the system would not know how to split a complex request into groups or which groups should depend on others.
Why Do Some Images Reference the Body Text?
Because an image group can use ref_group to reference the body text as creative context. This helps generated images match the topic more accurately.
Are Upload Groups Charged?
Upload groups do not call an AI generation model, so they usually do not incur model generation costs. However, template creation, web search, PE, or other AI-generated groups may still cost credits.
When Should I Retry a Group?
Retry a single group when the overall result is usable but one group is not satisfactory. For example, if the body text is good but the cover image is weak, retry only the cover image group.
Summary
GroupCombo turns complex creative requests into reusable templates. It first uses a blueprint to plan the structure, then generates independent PE for each content group, and finally lets users run the template repeatedly through scene variables.
Its best use case is not generating a one-time result, but continuously producing a category of content, such as social media posts, marketing image-text content, tutorials, brand asset packs, article illustrations, and multi-platform content templates.