Pre-paid credits.
Pay-per-call.

Premium cinematic text-to-image. Photoreal people, fashion, product.

Kiara Image Pro turns a text prompt into a photorealistic image. Tuned for portraits, fashion, beauty, product shots, and controlled scene direction. Accepts optional reference images for composition guidance.

Premium instruction-following image model. Clean edits + multi-reference.

Strong for accurate prompt details, product-style shots, clean edits, and multi-reference composition (up to 8 reference images). Best when prompt fidelity matters most.

Fast cinematic still model. Bold lighting, sharp prompt-driven ideas.

Good for stylized compositions, bold lighting, and sharp prompt-driven image ideas. 2K or 4K output. Accepts up to 4 reference images.

Flexible prompt + reference image generation. 2K or 4K.

Flexible prompt and reference-image generation. 2K or 4K output at the same price across both. Accepts up to 4 reference images.

Ultra-fast 8-step text-to-image with character/style locking.

Ultra-fast 8-step text-to-image. Best for rapid iteration and character/style locking when you need many candidates quickly.

Stylized creative image generation. 2K output, optional references.

Stylized creative image generation at 2K. Optional references — note: when you add a reference image the output size matches that image and the size picker is ignored. Strict content moderation; explicit references or outputs are refused at generation time.

Reference-image transformation. Put a subject into a new scene.

Reference-image transformation. One primary reference plus a scene prompt → a 2K or 4K composition. Best for 'put the subject in the reference into a new outfit / setting / scene'.

Skin texture + realism pass. Upload one image, pick an intensity.

Realism enhancement pass. Upload one image and pick an intensity — no prompt required. Adds skin texture and detail, then finishes up to 2K. Output matches the source orientation.

9:16 character portrait from a face reference + scene prompt.

Vertical character portrait. Upload one face reference and write a scene prompt. Photorealism and safety/quality anchors are locked on every generation; your prompt adds scene and pose intent on top.

Fast face swap. Body/scene image + face image.

Fast face swap. Upload a body/scene image and a face image; the model places the face onto the body. Output matches the body image dimensions. For a 4K finish use FaceSwap PRO.

Outfit swap. Dress your character in clothes from a second image.

Outfit swap. Upload your character (slot 1 — keeps face + hair) and an outfit reference (slot 2 — clothing transfers from here). The model dresses slot 1 in the clothes from slot 2 while preserving the original hair and identity. Output matches the character image dimensions.

Premium image enhancement. Sharpen, denoise, 2K/4K finish.

Premium image enhancement. Upload one image — the model sharpens, denoises, and lands a 2K or 4K finish with subtle film grain. No prompt; the source is the brief.

Scene composite. Edit subjects into a new scene from a prompt.

Scene composite. Drop in one reference (or up to three — character 1, character 2, optional pose anchor) and write the scene + action. With a single reference it edits that subject into the new scene; with multiple it composes them together. Source dimensions preserved with a 4K finish.

Reference-driven generation. Style/pose anchor + prompt.

Reference-driven generation. Stage one image as the style/pose anchor, write a prompt for the subject, and the pipeline keeps the reference context locked in. 4K finish.

Premium head swap with 4K finish.

Premium head swap with a 4K finish. Upload a body/scene image and a face reference; the model swaps the face onto the body and runs a 4K upscale. Use this when the output needs to be print-ready.

Scene redraw with auto-captioning. 4K finish.

Scene redraw with auto-captioning. Drop in a reference; the pipeline describes its pose, lighting, and outfit, applies a persona style, and renders the same scene at 4K. An optional descriptor adds style notes.

Cinematic light editing. Rewrite lighting on a finished image.

Cinematic light editing. Upload a finished image; the model preserves the subject and scene exactly but rewrites the lighting direction, intensity, and color. No prompt needed — just the source.

Highest-quality multimodal video. Text alone or guided by references.

Highest-quality multimodal video. Use text alone or guide the render with reference images and clips. Up to 2K, 1–15 seconds.

Fast multimodal video for quick iterations.

Fast multimodal video for quick iterations with optional reference images and clips. 1–15 seconds.

Fast multimodal video. Text alone or guided by references.

Fast multimodal video. Use text alone or guide the render with reference images and clips. 1–15 seconds.

Highest-quality image-to-video. Animate a start image.

Highest-quality image-to-video. Requires one start image and animates it from the prompt. 1–15 seconds.

Fast image-to-video, tuned for rapid previews.

Fast image-to-video. Requires one start image and is tuned for rapid previews. 1–15 seconds.

Highest-quality text-to-video. No reference needed.

Highest-quality text-to-video. No reference image required. 1–15 seconds.

Fast text-to-video for quick exploration.

Fast text-to-video. No reference image required, best for quick exploration. 1–15 seconds.

Reference-image motion model for adult character clips.

Reference-image motion model for adult character clips. Requires one image and keeps output vertical (9:16) at 720p.

Image-to-video. Your image is the first frame; prompt drives motion.

Grok Imagine Video 1.5 (preview). Image-to-video only — your image is the first frame and the prompt drives the motion. 480p or 720p, 1–15 seconds.