Image and Video Generation Models
This page summarizes the current image and video generation model families available on 4All API and explains the naming rules for model suffixes. The live list and pricing are subject to the console Model Marketplace.
Image Models
Section titled “Image Models”| Family | Representative Models | Description |
|---|---|---|
| GPT-Image | gpt-image-2、gpt-image-2-fast、gpt-image-1.5、gpt-image-1 | OpenAI image generation, -fast is the accelerated version |
| Gemini Images (Nano Banana class) | gemini-3.1-flash-image、gemini-3.0-pro-image、gemini-3-pro-image | Google image generation and editing |
| Imagen | imagen-4.0-generate-preview | Google Imagen text-to-image |
For usage, see GPT-Image Image Generation and Nano Banana Image Generation.
Image Model Suffixes
Section titled “Image Model Suffixes”You can specify aspect ratio and resolution by adding suffixes to the base model name, with no need to set them separately in parameters:
- Aspect ratio:
-portrait(portrait),-landscape(landscape),-square(square),-three-four(3:4),-four-three(4:3); no suffix means the default aspect ratio; - Resolution:
-2k,-4k(only GPT-Image supports 4k); - Suffixes can be combined, for example
gpt-image-2-fast-portrait-4k,gemini-3.1-flash-image-square-2k.
Video Models
Section titled “Video Models”Video generation uses an asynchronous task protocol (submit task → poll for result), see Video Generation (Async Protocol).
| Family | Representative Models | Description |
|---|---|---|
| VEO 3.1 | veo_3_1_t2v、veo_3_1_i2v_s、veo_3_1_r2v_fast | Google VEO, divided into standard / _fast / _lite / _s tiers, some support _1080p |
| Omni Flash | omni_flash_6s_portrait、omni_flash_i2v_10s_landscape | Fast video model, supports 4s / 6s / 8s / 10s durations |
| Abra | abra_t2v_8s_landscape、abra_r2v_10s_portrait | Supports 8s / 10s durations |
Video Model Naming Rules
Section titled “Video Model Naming Rules”Model names are composed as family + generation mode + duration + aspect ratio:
- Generation mode:
t2v(text-to-video),i2v(image-to-video),r2v(reference-image-to-video); when Omni Flash does not include a generation-mode segment, it defaults to text-to-video; - Duration:
4s/6s/8s/10s(supported ranges vary by family; see the Model Marketplace); - Aspect ratio:
portrait/landscape; - Tier / resolution (VEO only):
fast,lite,s,1080p.
Recommendation: use Omni Flash or VEO _lite / _fast for drafts and batch scenarios, and use the VEO standard tier when prioritizing final output quality.