Image, Audio, and Video Slash Commands

CodeBot can route media requests directly to the generation tools when a prompt starts with /image, /audio, or /video.

Use these commands when you already know that you want an image, spoken audio, or video. They skip the normal question-and-planning step and call the matching media tool directly.

Basic Image Generation

/image a cozy product photo of a red backpack on a wooden table

The text after /image becomes the image prompt. CodeBot calls generateImage and shows the generated image in the conversation.

Basic Audio Generation

/audio Hello from CodeBot!

The text after /audio becomes the text that will be spoken. CodeBot calls generateAudio and shows a player for the generated audio when the provider finishes.

Basic Video Generation

/video a six-second handheld shot of a red backpack on a wooden table

The text after /video becomes the video prompt. CodeBot calls generateVideo and shows the generated video when the provider finishes.

Fine-Tuning Your Request

Your prompt can include additional parameters and filenames to further describe the requested media generation beyond just plain text.

Save the Result

Add a target file name or path to save the generated media into the current workspace by including the name or path of the file you want to create. The file may not exist yet.

/image a red backpack on a wooden table backpack-v1.png
/audio hello from CodeBot welcome.mp3
/video slow orbit around the red backpack backpack-demo.mp4

If the file name is relative, Campfire saves it relative to the current workspace. You can also use an absolute path.

Supported image target extensions are png, jpg, jpeg, webp, and gif. Supported audio target extensions are mp3, wav, aac, flac, opus, and pcm. Supported video target extensions are mp4, m4v, mov, and webm.

Reference Images

Add an existing image path to use it as a visual reference, simply by specifying the name or path of an existing image.

/image create a new catalog image using References/backpack-front.png backpack-catalog-v1.png
/video animate this product with a gentle camera move References/backpack-front.png backpack-spin.mp4

Reference paths must point at existing local image files. Relative reference paths are resolved inside the current workspace. Note, not all models or workflows support reference images.

Image Arguments

/image understands the image-generation arguments that CodeBot passes to generateImage.

Argument Example Use
model= model=gpt-image-2 Use a specific image model or workflow model ID.
workflow= workflow=character-scene Shortcut for a ComfyUI workflow model.
variant= variant=kontext Add a workflow variant. Requires workflow= unless the value is a full workflow model ID.
seed= seed=12345 Use a deterministic seed when the provider supports it.
size= size=1536x1024 Request an image size.
quality= quality=high Request a provider-specific quality setting.
outputFormat= outputFormat=webp Request a specific output format.
background= background=transparent Request a background mode when supported.
negativePrompt= negativePrompt=blur Describe what the image should avoid. You can also use a negative: section in the prompt.
guidance= guidance=save-as-reference Tell the host how the result should be handled.
n= n=4 Request multiple images.
reference= reference=References/lina.png Add a reference image path.
source= source=References/source.png Use an existing image as the image-edit source.
mask= mask=References/mask.png Use an existing image as the edit mask.
save= save=Results/lina-v1.png Set the saved output path.

Arguments can appear before or after the prompt text.

/image workflow=character-scene seed=12345 size=1536x1024 Lina riding a mountain bike lina-bike-v1.png

Use quotes when an argument value contains spaces:

/image guidance="save as the new character reference" Lina portrait lina-reference.png

Audio Arguments

/audio understands the text-to-audio arguments that CodeBot passes to generateAudio.

Argument Example Use
model= model=gpt-4o-mini-tts Use a specific audio model.
voice= voice=alloy Select a voice name or provider-specific voice ID.
instructions= instructions="warm, calm, conversational" Control tone, pacing, emotion, pronunciation, accent, or delivery style.
style= style="radio announcer" Alias for instructions=.
language= language=es Provide a language or locale hint when supported.
locale= locale=es-MX Alias for language=.
format= format=mp3 Request an audio output format.
outputFormat= outputFormat=wav Alias for format=.
guidance= guidance=play-only Tell the host how the result should be handled.
save= save=Results/welcome.mp3 Set the saved output path.

The remaining text becomes the exact text to speak.

/audio voice=alloy format=mp3 instructions="warm, clear, friendly" Welcome to CodeBot welcome.mp3

Use quotes when the spoken text or instructions contain spaces that must stay together as one argument value:

/audio voice=Kore language=en-US instructions="soft, confident narration" "The build completed successfully."

If you omit the saved file name, Campfire uses audio.<format> in the current workspace, defaulting to audio.mp3.

Video Arguments

/video understands the video-generation arguments that CodeBot passes to generateVideo.

Argument Example Use
model= model=veo Use a specific video model or workflow model ID.
workflow= workflow=video-style Shortcut for a workflow model ID.
aspectRatio= aspectRatio=16:9 Request a video aspect ratio.
resolution= resolution=1080p Request a provider-specific resolution or quality.
durationSeconds= durationSeconds=6 Request a duration when supported.
negativePrompt= negativePrompt=blur Describe what the video should avoid.
guidance= guidance=save-as-draft Tell the host how the result should be handled.
reference= reference=References/backpack.png Add a reference image path.
save= save=Results/backpack-demo.mp4 Set the saved output path.
/video model=veo aspectRatio=16:9 durationSeconds=6 slow handheld shot of a red backpack backpack-demo.mp4

duration= is also accepted as an alias for durationSeconds=.

Save and Reference Aliases

These argument names all set the output path:

save=
saveTo=
suggestedFilePath=
output=

For audio, format= and outputFormat= both set the requested audio format.

These argument names all add a reference image:

reference=
referenceImage=
referenceFile=
referenceFilePath=

You can also provide bare existing image paths as references and bare target filenames as save paths.

Skills

Media slash commands can load Campfire skills.

Use skill=name:

/image skill=lina Lina riding a mountain bike lina-bike-v1.png

Or use the shorthand skill-name:

/image skill-lina Lina riding a mountain bike lina-bike-v1.png

For /image, the skill must be tagged for image or media. For /audio, the skill must be tagged for audio, speech, or media. For /video, the skill must be tagged for video or media. A skill only needs one matching media tag.

For image and video, skill prompt content is added before the user request. After the skill is expanded, Campfire parses the combined prompt for media arguments. This lets a skill provide defaults such as:

workflow=character-scene:kontext
seed=12345

If the command itself also provides the same argument, the command wins.

/image skill-lina workflow=character-scene:sdxl Lina in a forest lina-forest-v1.png

In this example, the workflow= in the command overrides any workflow supplied by the lina skill.

For audio, skill content is parsed for audio parameters, and the remaining skill text is added to instructions. It does not become spoken text.

/audio skill-narrator voice=alloy "The package is ready."

ComfyUI Workflows

For ComfyUI, workflow= is a shortcut for model=comfyui-workflows:....

These two commands are equivalent:

/image model=comfyui-workflows:character-scene Lina in a forest
/image workflow=character-scene Lina in a forest

Workflow variants use one more : segment:

/image workflow=character-scene:kontext Lina in a forest

The exact workflow and variant IDs come from listImageWorkflows.

Prompt Text

Arguments are removed before the media tool is called. For /image and /video, the remaining prompt should read like the creative instruction. For /audio, the remaining text is the exact text to speak. Arguments carry model, workflow, voice, format, seed, reference, size, and save-path choices.

When in doubt, put technical options as name=value arguments and keep the rest as plain prompt text.