Image Generation

Image generation — coming soon

Text-to-image and image-to-image diffusion on Gonka validators. Final model selection (SDXL Turbo, Flux Schnell/Pro, or a community fine-tune) still being benchmarked for throughput vs quality.

Expected tasks
- Text-to-image (prompt → PNG)
- Image-to-image (prompt + source → PNG)
- Inpainting (prompt + source + mask → PNG)
- Upscaling (4×, 8×)

Expected pricing: sub-$0.01 per image at 1024×1024. Competitive with Replicate, cheaper than OpenAI DALL·E.