Image to Prompt v1.0.0

AI & ML image-to-prompt 3 actions MCP

Turn images into detailed Flux/SD generation prompts with a local Ollama vision model. Runs entirely offline; batch a whole folder or describe a single image.

Open in Omniscient Back to apps

The omniscient:// link needs a protocol handler on your machine, which is not built yet. Until then it does nothing — use the commands below.

Actions

Jump to: describedescribe_folderlist_models

Describe image

describe

Generate a detailed image-generation prompt for one image using a local vision model.

Inputs

Name	Type	Required / default	Values and notes
`image`	path	required
`model`	string	optional, default `""`
`style`	enum	optional, default `flux`	`fluxsdshorttags`
`max_tokens`	int	optional, default `1200`

Outputs

Name	Type	Notes
`prompt`	text
`model`	string

Describe a folder

describe_folder

Generate prompts for every image in a folder and write them to a JSON file.

Inputs

Name	Type	Required / default	Values and notes
`folder`	path	required
`model`	string	optional, default `""`
`style`	enum	optional, default `flux`	`fluxsdshorttags`
`max_tokens`	int	optional, default `1200`
`out`	path	optional, default `""`

Outputs

Name	Type	Notes
`count`	int
`out_file`	path

List vision models

list_models

List the local Ollama vision models available for describing images.

Inputs

This action takes no inputs.

Outputs

Name	Type	Notes
`models`	json

Use it

Describe image

Command line

shell

python app.py --action describe --params '{"image": "D:/path/to/image"}'

Run from the app’s own folder. Single quotes work in bash and PowerShell; in cmd.exe use double quotes and escape the inner ones.

MCP tool

mcp

run_action("image-to-prompt", "describe", {"image": "D:/path/to/image"})

Workflow node

json

{
  "id": "n0",
  "app": "image-to-prompt",
  "action": "describe",
  "params": {
    "image": "D:/path/to/image"
  },
  "inputs": {}
}

Describe a folder

Command line

shell

python app.py --action describe_folder --params '{"folder": "D:/path/to/folder"}'

Run from the app’s own folder. Single quotes work in bash and PowerShell; in cmd.exe use double quotes and escape the inner ones.

MCP tool

mcp

run_action("image-to-prompt", "describe_folder", {"folder": "D:/path/to/folder"})

Workflow node

json

{
  "id": "n0",
  "app": "image-to-prompt",
  "action": "describe_folder",
  "params": {
    "folder": "D:/path/to/folder"
  },
  "inputs": {}
}

List vision models

Command line

shell

python app.py --action list_models --params '{}'

Run from the app’s own folder. Single quotes work in bash and PowerShell; in cmd.exe use double quotes and escape the inner ones.

MCP tool

mcp

run_action("image-to-prompt", "list_models", {})

Workflow node

json

{
  "id": "n0",
  "app": "image-to-prompt",
  "action": "list_models",
  "params": {},
  "inputs": {}
}

Example prompts

Things you can say to Claude, Cursor or Codex once the MCP server is registered.

describe this image as a Flux prompt
generate SD prompts for every image in D:/renders
give me short comma-separated tags for this photo