sourceful/riverflow-2.0-fast

Agentic image model optimized for high-quality, fast generations supporting font control

120 runs

Readme

Riverflow 2.0 Fast

Introduction

Riverflow 2.0 Fast is an agentic image model optimized for speed + quality, positioned as an upgrade vs flux-2-dev and qwen-image-2512 / qwen-image-edit-2511.

Note: While named “Fast”, the model is still quality-optimized and can take longer on complex requests.

Key features supported by Riverflow 2.0 Fast:

  • Text-to-image (t2i)
  • Image-to-image (i2i)
  • Font Control (text-in-image accuracy with specific fonts)
  • Transparency option

Riverflow 2.0 Fast — Text-to-image (t2i)

instruction: string

  • Min 2 chars
  • No strict max (10,000 chars is reasonable)

resolution: 1K | 2K

  • Default: 2K
  • No 4K support on Fast

aspectRatio:

auto | 21:9 | 16:9 | 3:2 | 4:3 | 5:4 | 1:1 | 4:5 | 3:4 | 2:3 | 9:16
- Default: auto

fontInputs: (see Font Control section below)

transparency: boolean
- Default: false

enhancePrompt: boolean
- Default: false

maxIterations: 1–3
- Default: 3

safetyChecker: boolean
- Default: true


Riverflow 2.0 Fast — Image-to-image (i2i)

Endpoint: i2i (see API spec playground)

imageUrls: up to 4 image URLs

  • URLs strongly preferred
  • Data URIs supported but constrained by request size limits

instruction: string

  • Min 2 chars
  • No strict max (10,000 chars is reasonable)

resolution: 1K | 2K

  • Default: 2K
  • No 4K support on Fast

aspectRatio:

auto | 21:9 | 16:9 | 3:2 | 4:3 | 5:4 | 1:1 | 4:5 | 3:4 | 2:3 | 9:16
- Default: auto

fontInputs: (see Font Control section below)

transparency: boolean

  • Default: false

maxIterations: 1–3

  • Default: 3

enhancePrompt: boolean

  • Default: false

safetyChecker: boolean

  • Default: true

Font Control (Fast)

Font control enables users to generate clean, legible text-in-image using specific fonts, without external editing tools.

How it works

  • Supports up to 2 fonts
  • Each font requires:
  • A font URL (.ttf, .otf, .woff, .woff2)
  • The exact text (up to ~300 characters) that will appear in the image
    • This allows the model to load the relevant glyph subset into working memory

Example prompt pattern

In instruction, a user might write:

On-image copy (must be perfectly legible, uppercase, centered):
“THE LATE-NIGHT STACK”
Sub-line (smaller, letter-spaced, below):
“HOT. FAST. UNREASONABLE.”
Font: Kenia.

Then in fontInputs, they should provide: - the Kenia font URL - the exact text again: HOT. FAST. UNREASONABLE. (and/or the headline)


Getting Costs

Costs are known at request time for standard generations.

Fast pricing

  • Fast 1K: $0.02 / image
  • Fast 2K: $0.04 / image
  • No 4K support

Typical latency: - ~15 seconds per image - Can be longer for challenging requests

Font Control pricing

  • +$0.03 per font (max 2)
Model created
Model updated