qwen-image/qwen-image-edit-2511/lora
LoRA-based visual editing model offering structure-aware asset transformation for creative pros
Nano Banana 2 (Gemini 3.1 Flash Image) is Google DeepMind’s Flash-tier image generation model designed for high-speed, instruction-following visual creation with strong typography rendering and real-world knowledge integration.
Nano Banana 2 text to image converts a single text prompt into 1–4 still images per request, supporting reproducible generation via seed control and flexible resolution tiers from 0.5K to 4K. It is optimized for fast iteration, predictable framing, and production-ready outputs suitable for marketing visuals, product mockups, social media assets, and storyboards.
Output format: png, jpeg, or webp. Outputs: still images (batch size: 1–4).
The following controls are exposed for Nano Banana 2 Text-to-Image.
| Parameter | Required | Type | Default | Range / Options | Description |
|---|---|---|---|---|---|
| prompt* | Yes (*) | string (str) | A cinematic close-up portrait of an American woman standing under neon lights in rainy Tokyo... | — | Text prompt describing subject, scene, lighting, style, and composition. Be clear and structured for best results. |
| num_images | No | integer | 1 | 1–4 | Number of images to generate per request. Use multiple outputs for variation exploration. |
| seed | No | integer | 0 | Any integer | Controls randomness. Use the same seed to reproduce similar results; change it for new variations. |
| aspect_ratio | No | string | auto | auto, 21:9, 16:9, 3:2, 4:3, 5:4, 1:1, 4:5, 3:4, 2:3, 9:16 | Output framing. "auto" preserves natural composition; select a ratio for specific layout needs. |
| resolution | No | string | 1K | 0.5K, 1K, 2K, 4K | Target resolution. Higher tiers increase detail and cost. |
| output_format | No | string | png | jpeg, png, webp | Export format. PNG is ideal for text clarity; JPEG/WEBP reduce file size. |
| safety_tolerance | No | integer | 4 | 1–6 | Content moderation strictness. 1 is strictest; 6 is most permissive. |
| limit_generations | No | boolean | true | true / false | If enabled, limits each prompt round to one generation for controlled iteration. |
| enable_web_search | No | boolean | false | true / false | Allows integration of recent web information to improve factual accuracy. |
Use lower resolution with multiple seeds during ideation, then upscale to higher resolution once composition is finalized.
1) Write a structured prompt: Subject → action → environment → style → camera/lighting.
2) Set aspect_ratio to match your final deliverable (or keep "auto" for natural framing).
3) Choose resolution based on draft vs. production needs.
4) Set seed if you require reproducible outputs.
5) Generate 1–4 images and review composition, lighting, and typography.
6) Iterate by adjusting small variables (pose, color, mood) rather than rewriting the entire prompt.
LoRA-based visual editing model offering structure-aware asset transformation for creative pros
High-speed model for rapid text-to-image creation with rich detail and flexible format control.
Transform visuals with smart region edits and multi-image blending for precise, high-fidelity results.
Generate accurate brand visuals with high-fidelity text-to-image control.
Create consistent visual stories with advanced image editing and multi-scene control.
Turn written concepts into detailed visuals with precise image synthesis for creative teams.
Nano Banana 2 text-to-image is designed for fast iteration and consistent prompt-following. It’s a great fit for rapid concept exploration, marketing drafts, thumbnails, and generating multiple variations quickly.
Nano Banana 2 text-to-image supports common aspect ratios including 21:9, 16:9, 3:2, 4:3, 5:4, 1:1, 4:5, 3:4, 2:3, and 9:16. Choose the ratio that matches your target layout (banner, square post, story, etc.).
Nano Banana 2 text-to-image supports generating 1–4 images per request via the “Number of Images” parameter. For more variety, keep the same prompt and rerun with different seeds (when available) or small prompt variations.
safety_tolerance controls how strict content moderation is. Lower values are stricter; higher values are more permissive. If you’re generating brand-safe or public-facing content, use a stricter setting.
If enabled, enhance_prompt tries to expand or refine your prompt to improve descriptiveness and coherence. If you prefer precise control, keep it off and write explicit constraints (subject, style, lighting, composition) yourself.
Nano Banana 2 text-to-image can output images in jpeg, png, or webp. Use png for crisp graphics and text, jpeg for smaller files, and webp for a good quality-size balance.
RunComfy is the premier ComfyUI platform, offering ComfyUI online environment and services, along with ComfyUI workflows featuring stunning visuals. RunComfy also provides AI Models, enabling artists to harness the latest AI tools to create incredible art.