Ideogram Image Generation Compared: How to Choose Among Six Core Models

Introduction: The Ideogram Editing Matrix

Ideogram has built a product line around character consistency and flexible image editing—from pure text-to-image to masked local edits, scene remixes, and identity-locked character workflows. On FuseAI Tools, six named modes cover the full creative path; the question is which lever to pull for each job.

This guide maps character-edit, character-remix, character, v3-text-to-image, v3-edit, and v3-remix to representative API fields and the exact routes where you can run them.

Ideogram hub: /home/ideogram

I. Snapshot: Six Models at a Glance

Model Core function Inputs Mask Character ref Positioning
character-editMasked character editimage + mask + ref + text✅ (1)Local character fix
character-remixCharacter scene remiximage + ref + text✅ (1)New background, same face
characterCharacter in new sceneref + text✅ (1)Scene from reference only
v3-text-to-imageText-to-imagetextCreate from zero
v3-editInpainting / local editimage + mask + textGeneric masked fill
v3-remixGlobal remiximage + textWhole-image restyle

II. Six Models Deep Dive

Model 1: character-edit

Edit a masked region on a base portrait while locking identity to a separate reference still—ideal for expression tweaks, wardrobe details, or localized refinements.

{
  "model": "ideogram-character-edit",
  "prompt": "Fill the masked region (max 5000 characters)",
  "imageUrl": "Base image URL",
  "maskUrl": "Mask URL (white = edit, black = keep)",
  "referenceImageUrls": ["Character reference (1 image)"],
  "renderingSpeed": "BALANCED",
  "expandPrompt": true,
  "numImages": "1",
  "seed": 0
}

Mechanics: upload base image, mask (same dimensions), and one character reference. FuseAI Tools accepts JPG/PNG/WEBP up to 10MB per asset.

Best for: eye/expression changes, localized outfit edits, identity-safe inpainting.

Route: /home/ideogram/character-edit

Model 2: character-remix

Keep a character recognizable while remixing scene and context from an existing frame plus prompt. strength balances fidelity to the source image vs creative freedom.

{
  "model": "ideogram-character-remix",
  "prompt": "Scene remix brief (max 5000 characters)",
  "imageUrl": "Base image URL",
  "referenceImageUrls": ["Character reference (1 image)"],
  "renderingSpeed": "BALANCED",
  "style": "AUTO",
  "expandPrompt": true,
  "imageSize": "square_hd",
  "numImages": "1",
  "strength": 0.8,
  "negativePrompt": "",
  "seed": 0
}

Strength guide: higher (0.8–1.0) preserves more of the original frame; lower (0.1–0.5) follows the prompt more aggressively.

Best for: background swaps, environmental storytelling, “same hero, new location.”

Route: /home/ideogram/character-remix

Model 3: character

Place a character into a brand-new scene using only a reference still and text—no base image required.

{
  "model": "ideogram-character",
  "prompt": "Scene description (max 5000 characters)",
  "referenceImageUrls": ["Character reference (1 image)"],
  "renderingSpeed": "BALANCED",
  "style": "AUTO",
  "expandPrompt": true,
  "imageSize": "square_hd",
  "numImages": "1",
  "negativePrompt": "",
  "seed": 0
}

vs character-remix:

Contrast character character-remix
Base imageNot requiredRequired
Scene sourcePrompt onlyOriginal frame + prompt
strengthN/A

Best for: character posters, multi-scene campaigns, net-new compositions anchored to one face.

Route: /home/ideogram/character

Model 4: v3-text-to-image

General-purpose generation from text with style, speed, size, MagicPrompt expansion, and optional negative prompt.

{
  "model": "ideogram-v3-text-to-image",
  "prompt": "Image description (max 5000 characters)",
  "renderingSpeed": "BALANCED",
  "style": "AUTO",
  "expandPrompt": true,
  "imageSize": "square_hd",
  "negativePrompt": "",
  "seed": 0
}

Style options: AUTO, GENERAL, REALISTIC, DESIGN.

Rendering speed: TURBO (fastest), BALANCED (default), QUALITY (slowest, highest fidelity).

Image size examples: square, square_hd, portrait_4_3, portrait_16_9, landscape_4_3, landscape_16_9.

Best for: concept art, product sketches, marketing stills without a reference photo.

Route: /home/ideogram/v3-text-to-image

Model 5: v3-edit

Mask-driven inpainting without a character reference—swap objects, fix backgrounds, or remove artifacts in a selected region.

{
  "model": "ideogram-v3-edit",
  "prompt": "What to paint inside the mask (max 5000 characters)",
  "imageUrl": "Source image URL",
  "maskUrl": "Mask URL",
  "renderingSpeed": "BALANCED",
  "expandPrompt": true,
  "seed": 0
}

vs character-edit: v3-edit is identity-agnostic; character-edit adds referenceImageUrls so faces stay on-model.

Best for: prop swaps, watermark cleanup, localized retouching on non-character assets.

Route: /home/ideogram/v3-edit

Model 6: v3-remix

Transform an entire image toward a new style or content direction—no mask, no character reference, with adjustable strength.

{
  "model": "ideogram-v3-remix",
  "prompt": "Remix direction (max 5000 characters)",
  "imageUrl": "Source image URL",
  "renderingSpeed": "BALANCED",
  "style": "AUTO",
  "expandPrompt": true,
  "imageSize": "square_hd",
  "numImages": "1",
  "strength": 0.8,
  "negativePrompt": "",
  "seed": 0
}

vs character-remix: v3-remix does not take a character reference; use character-remix when facial consistency matters.

Best for: material changes, illustration styles, holistic art-direction shifts.

Route: /home/ideogram/v3-remix

III. Cross-Model Comparison

3.1 Capability matrix

Feature char-edit char-remix char v3 T2I v3-edit v3-remix
Text-to-image
Mask edit
Character reference
Global remix
strength
negativePrompt
imageSize

3.2 Character line vs V3 line

Product line Models Core idea
Charactercharacter-edit, character-remix, characterIdentity-locked edits and scene work
V3 createv3-text-to-imagePrompt-only generation
V3 editv3-edit, v3-remixGeneric masked or global edits

IV. Selection Decision Trees

Character workflows

Character consistency required?
|
|-- Local change inside a region (eyes, outfit detail)
|   `-- character-edit (+ mask)
|
|-- New environment but same person
|   |-- Have a frame to remix -> character-remix
|   `-- Start from reference only -> character
|
`-- Campaign with many scenes
    `-- character for new shots; character-remix to adapt existing frames

General image workflows

No character reference?
|
|-- Create from scratch -> v3-text-to-image
|
`-- Edit existing photo
    |-- Local swap / cleanup -> v3-edit (+ mask)
    `-- Whole-image restyle -> v3-remix

V. Practical Payload Examples

Expression tweak (character-edit)

{
  "model": "ideogram-character-edit",
  "prompt": "Open the eyes and add a gentle forward-looking smile",
  "imageUrl": "https://example.com/portrait.jpg",
  "maskUrl": "https://example.com/eyes-mask.png",
  "referenceImageUrls": ["https://example.com/character-ref.jpg"]
}

Background swap (character-remix)

{
  "model": "ideogram-character-remix",
  "prompt": "Place the subject at Shibuya crossing at night, neon reflections",
  "imageUrl": "https://example.com/character.jpg",
  "referenceImageUrls": ["https://example.com/character-ref.jpg"],
  "strength": 0.7
}

Landscape text-to-image

{
  "model": "ideogram-v3-text-to-image",
  "prompt": "Cinematic dusk lakeside, reeds and lily pads with golden reflections",
  "style": "REALISTIC",
  "renderingSpeed": "QUALITY",
  "imageSize": "landscape_16_9"
}

Object swap (v3-edit)

{
  "model": "ideogram-v3-edit",
  "prompt": "Replace the coffee cup with a red mug",
  "imageUrl": "https://example.com/room.jpg",
  "maskUrl": "https://example.com/cup-mask.png"
}

VI. Final Recommendations

Use case Model Why
Identity-safe local editcharacter-editMask + reference lock the face
Same character, new backdropcharacter-remixRemix with strength control
Character poster from refcharacterNo base frame required
Fresh illustration / photov3-text-to-imageFull V3 style and size stack
Masked object replacev3-editGeneral inpainting
Global style shiftv3-remixWhole-image transformation

One-line playbook:

Open every Ideogram mode from /home/ideogram.