Images and web search

Image generation and web research are separate from the text model. You can combine providers—for example, Ollama for text, Pexels for images, and a private SearXNG instance for search.

Image providers

Set IMAGE_PROVIDER to one of the supported values.

Capability	`IMAGE_PROVIDER`	Required or related settings
Pexels stock images	`pexels`	`PEXELS_API_KEY`
Pixabay stock images	`pixabay`	`PIXABAY_API_KEY`
Gemini Flash images	`gemini_flash`	`GOOGLE_API_KEY`
Nano Banana Pro	`nanobanana_pro`	`GOOGLE_API_KEY`
DALL·E 3	`dall-e-3`	`OPENAI_API_KEY`, optional quality
GPT Image 1.5	`gpt-image-1.5`	`OPENAI_API_KEY`, optional quality
ComfyUI	`comfyui`	`COMFYUI_URL`, optional workflow
Open WebUI	`open_webui`	`OPEN_WEBUI_IMAGE_URL`, `OPEN_WEBUI_IMAGE_API_KEY`
OpenAI-compatible image API	`openai_compatible`	Base URL, API key, and model

Set DISABLE_IMAGE_GENERATION=true to create presentations without generated or searched images.

Stock-image example

IMAGE_PROVIDER=pexels
PEXELS_API_KEY=replace-with-your-key

OpenAI image example

IMAGE_PROVIDER=gpt-image-1.5
OPENAI_API_KEY=replace-with-your-key
GPT_IMAGE_1_5_QUALITY=medium

Supported GPT Image 1.5 quality values are low, medium, and high; the default is medium. DALL·E 3 supports standard and hd quality.

ComfyUI example

IMAGE_PROVIDER=comfyui
COMFYUI_URL=http://host.docker.internal:8188
COMFYUI_WORKFLOW={"6":{"inputs":{"text":"placeholder prompt"},"class_type":"CLIPTextEncode","_meta":{"title":"Input Prompt"}}}

COMFYUI_WORKFLOW contains the API-exported workflow JSON. The prompt node must be titled Input Prompt so Presenton can inject each generated prompt.

Configure ComfyUI step by step

Expose the server safely, prepare the prompt node, export the API workflow, and troubleshoot generation.

OpenAI-compatible image API

IMAGE_PROVIDER=openai_compatible
OPENAI_COMPAT_IMAGE_BASE_URL=https://proxy.example.com/v1
OPENAI_COMPAT_IMAGE_API_KEY=replace-with-your-key
OPENAI_COMPAT_IMAGE_MODEL=gpt-image-1

This is useful for gateways such as LiteLLM or another service that exposes a compatible image endpoint. It does not change the text provider.

Web search

Web search is opt-in. Enable it only when decks need current public information.

WEB_GROUNDING=true
WEB_SEARCH_PROVIDER=auto
WEB_SEARCH_MAX_RESULTS=5

`WEB_SEARCH_PROVIDER`	Behavior	Additional setting
`auto`	Uses native search for supported OpenAI, Google, or Anthropic configurations; otherwise remains off	None
`native`	Requests native provider search	A compatible text provider
`searxng`	Uses a self-hosted SearXNG instance	`SEARXNG_BASE_URL`
`tavily`	Uses Tavily	`TAVILY_API_KEY`
`exa`	Uses Exa	`EXA_API_KEY`

WEB_SEARCH_MAX_RESULTS defaults to 5 and supports up to 10 results.

SearXNG
Tavily
Exa

WEB_GROUNDING=true
WEB_SEARCH_PROVIDER=searxng
SEARXNG_BASE_URL=http://searxng:8080
WEB_SEARCH_MAX_RESULTS=5

WEB_GROUNDING=true
WEB_SEARCH_PROVIDER=tavily
TAVILY_API_KEY=replace-with-your-key
WEB_SEARCH_MAX_RESULTS=5

WEB_GROUNDING=true
WEB_SEARCH_PROVIDER=exa
EXA_API_KEY=replace-with-your-key
WEB_SEARCH_MAX_RESULTS=5

Web results and generated images can introduce inaccurate, outdated, or licensed material. Review facts, citations, image rights, and attribution requirements before sharing a presentation.

Get started

Learn the basics

Features

Tutorials

API guides

Deploy and configure

Develop and contribute

This release

Image providers

Stock-image example

OpenAI image example

ComfyUI example

Configure ComfyUI step by step

OpenAI-compatible image API

Web search

​Image providers

​Stock-image example

​OpenAI image example

​ComfyUI example

Configure ComfyUI step by step

​OpenAI-compatible image API

​Web search

Image providers

Stock-image example

OpenAI image example

ComfyUI example

OpenAI-compatible image API

Web search