Full image creation workflow using Hermes Image Studio. Guides the agent through prompt engineering, preset selection, generation, upscaling, and batch exploration.
Installs into .claude/skills of the current project.
Are you the author of image-studio-workflow?
Add the live security badge to your README. It updates with every re-scan.
[](https://www.skillsdirectory.com/skills/theheavenlyd3mon-image-studio-workflow)
---
name: image-studio-workflow
description: "Full image creation workflow using Hermes Image Studio. Guides the agent through prompt engineering, preset selection, generation, upscaling, and batch exploration."
---
# Image Studio: Full Workflow
End-to-end skill for producing publish-ready images using the Hermes
Image Studio plugin (tools: `image_studio_presets`,
`image_studio_generate`, `image_studio_edit`, `image_studio_inpaint`,
`image_studio_animate`, `image_studio_batch`,
`image_studio_upscale`, `image_studio_gallery`,
`image_studio_save_prompt`, `image_studio_load_prompt`,
`image_studio_prompts`, `image_studio_history`).
## The Pipeline
### Phase 1: Discovery
Start with `image_studio_presets` if the user hasn't picked a style.
Discuss:
- What is the image for? (Twitter, article header, personal project)
- What mood or era fits the subject?
- Would a specific preset serve the purpose?
### Phase 2: Generate
Write a specific, detailed prompt. Generic prompts produce generic
results. Include:
- Subject and action
- Setting and environment
- Lighting and time of day
- Colors and mood
- Composition (wide shot, close-up, portrait)
Call `image_studio_generate` with the prompt and chosen preset.
### Phase 3: Evaluate (Optional)
If the result needs refinement:
- Change the preset for a different visual style
- Add more detail to the prompt for specificity
- Use a specific seed to lock composition while changing the prompt
- Add a negative prompt to exclude unwanted elements ("no people, no text")
- Use `image_studio_edit` to transform the image with a new prompt
- Use `image_studio_inpaint` with a mask to change only one region
### Phase 4: Batch (Optional)
If the user wants options, run `image_studio_batch` with count=4.
Present all variants and pick the strongest direction.
### Phase 5: Upscale (Final Step)
For the final selected image, run `image_studio_upscale` with the
image URL from the generation. This produces a 2x higher resolution
version suitable for Twitter, printing, or publication.
### Phase 5b: Animate (Optional)
If the user wants motion from a still:
- **If BFL/FLUX 3 tools are available** (`bfl_flux3_image_to_video`,
gated on Nous Portal paid sign-in), prefer them for highest quality.
The opening frame is the source image itself; polling and download are
handled by `bfl_flux3_get_result`.
- **Otherwise** run `image_studio_animate` with the image URL and a
motion prompt. The MP4 is saved to the `videos/` subfolder via FAL
(Kling v3 Pro or Veo 3.1 Fast).
### Phase 6: Deliver
Confirm the final file path and seed so the user can reference or
re-generate later. The image is already saved to the configured output
directory.
## Quick Reference
```
Step Tool When
----- ---- ----
Pick a preset image_studio_presets Always start here
Generate image_studio_generate One shot
Edit image_studio_edit Transform an existing image
Inpaint image_studio_inpaint Change only a masked region
Explore variants image_studio_batch count=4 User wants options
Upscale final image_studio_upscale Before publishing
Animate bfl_flux3_image_to_video* Premium motion (if available)
image_studio_animate FAL motion (Kling/Veo)
Browse history image_studio_history Re-find or tweak past work
Build gallery image_studio_gallery Show everything at once
* bfl_flux3_* needs Nous Portal paid sign-in; otherwise use image_studio_animate
```
## Output Location
All images are saved to `/Volumes/Spare Drive/Personal Stuff/Image Studio/`
by default, organized by date in the filename:
`YYYYMMDD_HHMMSS_preset_seed_subject.png`