Riverflow 2.0 Pro (Riverflow Pro)

Riverflow 2.0 Pro is a high-quality, agentic image generation and editing model designed for reliability, strong prompt-following, and autonomous self-correction. It is optimized for quality over speed and may take longer to produce the best result.

Recommended timeout: up to 10 minutes per request (longer reasoning = better outputs)

Model capabilities

1) Text-to-Image (T2I)

Generate images from a text instruction, with support for:

a) Multiple resolutions (1K, 2K, 4K) b) 10+ aspect ratios (including auto) c) Optional transparent backgrounds d) Optional prompt enhancement e) Multi-iteration agentic refinement for higher reliability f) Font Control for accurate, legible text rendering

2) Image-to-Image (I2I)

Edit or transform one or more input images using a text instruction. Common use cases:

a) Style transfer and re-rendering b) Background changes c) Object edits (add/remove/replace) d) Layout/scene changes e) Brand-safe variations f) Font Control for editing or adding text in-image g) Optional transparency h) Reference-Based Super-Resolution (detail fixing with reference images)

3) Font Control (Text accuracy in images)

Font Control improves accuracy and legibility of text rendered inside generated or edited images.

Key points

• Works for both text-to-image and image-to-image
• Supports up to 2 fonts
• Each font includes:
• a font file URL (.ttf, .otf, .woff, .woff2)
• the exact text to be rendered (up to ~300 characters per font)

Best practices In your instruction, explicitly specify:

1. the exact text (verbatim)
2. casing (e.g. ALL CAPS)
3. placement (e.g. centered, top-left)
4. legibility requirements (e.g. “must be perfectly legible”)
5. the font name
6. In fontInputs, provide the same text again to improve accuracy.