ComfyUI Node

Ideogram 3.0 Remix

Ideogram 3.0 Remix

By Runware·Created 2 years ago·Updated about a month ago· 140
Ideogram 3.0 Remix
  • referenceImages
  • advancedFeatures.watermark.image
  • providerSettings.ideogram.characterReferenceImages
  • providerSettings.ideogram.characterReferenceImagesMask
  • providerSettings.ideogram.styleReferenceImages
  • image
positivePrompt
negativePrompt
size1K (1:1)
numberResults1
advancedFeaturesfalse
advancedFeatures.watermark.bgColor
advancedFeatures.watermark.displayPosition(default)
advancedFeatures.watermark.fontColor
advancedFeatures.watermark.opacityfalse
advancedFeatures.watermark.opacity_value0.10
advancedFeatures.watermark.text
providerSettings.ideogram.magicPromptAUTO
providerSettings.ideogram.remixStrength50
providerSettings.ideogram.renderingSpeedDEFAULT
safetyfalse
safety.checkContentfalse
providerSettings.ideogram.styleCode
providerSettings.ideogram.stylePreset(default)
providerSettings.ideogram.styleTypeAUTO
ttlfalse
ttl_value60
outputFormatJPG
outputQuality95

Image-to-image is where "trust the prompt" stops working and "trust the strength slider" takes over. Ideogram 3.0 Remix is the hosted Ideogram API's take on that: you feed it reference images, a prompt, and a remix strength, and it transforms the source rather than copying it. It runs in the cloud through Runware, so there's no checkpoint to download and no VRAM to watch.

The 3.0 Remix is the version you want when you have some control and want to keep it. Unlike a blind txt2img pass, your reference images anchor composition and subject - the model is reworking what's there, not ignoring it. Think "make my product shot into a moody studio photo," not "new image, same seed."

How it works

Required inputs: referenceImages (one or more IMAGE tensors), positivePrompt, and width/height. The tensor goes up to Runware over REST, and the result comes back as a native IMAGE.

The knob that matters most:

  • providerSettings.ideogram.remixStrength - 1 to 100, default 50. Low values barely disturb the source; high values rewrite it dramatically. This is your fidelity dial. Start at 50, nudge one way or the other based on whether you're losing the subject or the prompt's intent.

Also worth a look: negativePrompt to push the model away from things, stylePreset (the 63-item art-direction list), and the character-reference sockets (characterReferenceImages + its mask) if you're keeping a consistent face across shots. magicPrompt on AUTO quietly rewrites your prompt into something the model likes better - good for beginners, mildly annoying if you've been precise on purpose. numberResults gives you up to 20 seeded variations, which is the cheapest way to explore since each one is fractions of a cent.

Note the character-reference masks come as IMAGE, not MASK - this pack expects them as image tensors, so keep that in mind when wiring.

Installing it

It's part of Runware's official pack, so:

cd ComfyUI/custom_nodes
git clone https://github.com/Runware/ComfyUI-Runware
pip install -r ComfyUI-Runware/requirements.txt

Restart, then drop your key into ComfyUI Settings → Runware API key (or RUNWARE_API_KEY). Dependencies are runware-sdk, pillow, soundfile - nothing heavy, no model files.

Where people get burned

The remix strength is the classic trap - people set it at 100 expecting "more remix" and get a painting that barely resembles the source. Treat it like a denoise slider: 30–50 for recognizable edits, 70+ when you genuinely want a new image that merely rhymes with the old one. And remember this is a paid API on a top-up account (the $20 minimum and mandatory card linking are the community's standing complaints, not the per-image price). Queue a numberResults of 2–4, look at what comes back, then go wider.

CategoryRunware/Image/ideogram

Inputs (28)

NameTypeDefaultDescription
referenceImagesIMAGE
positivePromptSTRINGText prompt describing elements to include in the generated output.
advancedFeatures.watermark.imageoptIMAGE
providerSettings.ideogram.characterReferenceImagesoptIMAGE
providerSettings.ideogram.characterReferenceImagesMaskoptIMAGE
providerSettings.ideogram.styleReferenceImagesoptIMAGE
negativePromptoptSTRINGPrompt to guide what to exclude from generation. Ignored when guidance is disabled (CFGScale ≤ 1).
sizeoptCOMBO1K (1:1)Output size. Only one of resolution or dimensions is sent.
numberResultsoptINT11–20Number of results to generate. Each result uses a different seed, producing variations of the same parameters.
advancedFeaturesoptBOOLEANfalseEnable to set advancedFeatures. Off uses the model's default.
advancedFeatures.watermark.bgColoroptSTRINGBackground color in hex format.
advancedFeatures.watermark.displayPositionoptCOMBO(default)Watermark position.
advancedFeatures.watermark.fontColoroptSTRINGText color in hex format.
advancedFeatures.watermark.opacityoptBOOLEANfalseEnable to set advancedFeatures.watermark.opacity. Off uses the model's default.
advancedFeatures.watermark.opacity_valueoptFLOAT0.100.1–1Watermark opacity.
advancedFeatures.watermark.textoptSTRINGWatermark text.
providerSettings.ideogram.magicPromptoptCOMBOAUTOAutomatic prompt enhancement.
providerSettings.ideogram.remixStrengthoptINT501–100Transformation intensity for Remix models. Higher values create more dramatic changes.
providerSettings.ideogram.renderingSpeedoptCOMBODEFAULTGeneration speed/quality tradeoff.
safetyoptBOOLEANfalseEnable to set safety. Off uses the model's default.
safety.checkContentoptBOOLEANfalseEnable or disable content safety checking.
providerSettings.ideogram.styleCodeoptSTRINGPredefined style code (8-char hex).
providerSettings.ideogram.stylePresetoptCOMBO(default)Artistic style preset.
providerSettings.ideogram.styleTypeoptCOMBOAUTOVisual style and rendering approach.
ttloptBOOLEANfalseEnable to set ttl. Off uses the model's default.
ttl_valueoptINT60Time-to-live (TTL) in seconds for generated content. Only applies when `outputType` is `URL`.
outputFormatoptCOMBOJPGFile format for the generated image.
outputQualityoptINT9520–99Compression quality of the output. Higher values preserve quality but increase file size.

Outputs (1)

NameTypeDescription
imageIMAGE