PayWeave Replicate

Fastest FLUX. ~2s per image. Cheap drafts and high-volume gen.

FLUX Dev (open weights). Better quality than schnell, slower.

BFL flagship. Best quality/speed balance for production work.

BFL flagship at 4MP output. For posters, prints, and high-detail use.

Black Forest Labs FLUX 2 Max. Highest-fidelity FLUX. Multi-image conditioning (up to 4 refs) and ≤2 MP output through this route.

Black Forest Labs FLUX 2 Pro. High-quality generation and editing with up to 4 reference images, ≤2 MP output through this route.

Black Forest Labs FLUX 2 Flex. Max-quality with tunable steps/guidance and up to 4 reference images, ≤2 MP output through this route.

Black Forest Labs FLUX 2 Klein 4B. Sub-second distilled FLUX. Cheap-and-fast tier; ≤2 MP output and ≤4 reference images through this route.

ByteDance Seedream 4.5. Strong spatial reasoning and world knowledge. 2K/4K output, up to 14 reference images for multi-image conditioning.

ByteDance Seedream 5.0 Lite. Image gen with built-in reasoning and example-based editing. 2K/3K output, up to 14 reference images.

ByteDance Seedream 4. Unified text-to-image + single-sentence editing at up to 4K. Up to 10 reference images.

ByteDance Seedream 3. Text-to-image at native 2K. Tight guidance control via guidance_scale.

xAI Grok Imagine. SOTA image generation and single-image editing from xAI.

Recraft V4. Latest Recraft model — strong prompt accuracy, art-directed composition, integrated text rendering.

Recraft V4 SVG. Production-ready vector output — clean geometry, structured layers, editable paths. Best for logos and icons.

Recraft V4 Pro. ~2048px output with V4 design taste. Print-ready and large-scale work.

Best-in-class for typography, posters, brand assets, and vector output.

Ideogram V3 turbo tier. Fast text rendering for logos and signage.

Ideogram V3 balanced tier. Better detail than turbo.

Ideogram V3 quality tier. Premium output for book covers and prints.

Google Imagen 4 fast tier. Drafts and rapid iteration.

Google Imagen 4 standard. Production-quality general image gen. Supports 1K/2K output.

Google Imagen 4 ultra. Skin texture, fabric weave, fine detail. Supports 1K/2K output.

Google Nano Banana 2 (Gemini 3.1 Flash Image) at 1K. Fast text-to-image and conversational editing; accepts up to 14 reference images.

Nano Banana 2 at 2K. Same model, higher-resolution output for print previews and social hero images.

Nano Banana 2 at 4K. Premium tier for posters, prints, and detail-critical assets.

Google Nano Banana Pro (Gemini 3 Pro Image) at 2K. Premium text-to-image and editing with strong text rendering and multi-subject consistency.

Nano Banana Pro at 4K. Highest-fidelity tier — posters, prints, and detail-critical hero imagery.

Stability AI SD 3.5 Large. Strong stylized output and full negative-prompt support.

SD 3.5 Large Turbo. 1-4 step variant, ~3× faster than SD 3.5 Large.

OpenAI GPT Image 1.5 (low quality). Cheapest tier with transparent-background support. Strong instruction following and text rendering.

GPT Image 1.5 (medium quality). Production sweet spot — supports transparent backgrounds.

GPT Image 1.5 (high quality, equivalent to `auto`). Highest fidelity tier with transparent-background support.

OpenAI GPT Image 2 (low quality). Cheapest tier — drafts, thumbnails, rapid iteration. Strong instruction following and sharp text rendering.

GPT Image 2 (medium quality). Sweet spot for production marketing assets.

GPT Image 2 (high quality, equivalent to `auto`). Highest fidelity — covers, hero imagery, dense text scenes.

Instruction-based image editing. "Change the background to a snowy mountain."

Inpainting with mask. Replace a masked region while preserving the rest.

Premium kontext editing. Higher fidelity for complex multi-region edits.

High-detail creative upscaler. Adds plausible texture, not a strict reconstruction.

Classic ESRGAN super-resolution. Fast, deterministic, no creative additions.

Face restoration for old or low-quality portraits.

Background removal. Returns the input image with transparent background.

Fast English TTS with 30+ voices. Compact 82M param model — great for high-volume narration.

Voice cloning TTS from a 5-30 sec reference clip. 17 languages supported.

Meta MusicGen. Generate music from text prompt; optionally condition on a melody clip.

Visual question answering. Pass an image and a prompt, get a text description or answer.