# Models

> The hosted lineup, by purpose.

You choose an alias by purpose; the service maps it to the model below. Raw model ids are not accepted. The lineup never includes a model whose terms bar commercial use of its output; some terms are still marked unclear, see [Licences](https://vincevangen.com/docs/licences.md). Prices are in credits on the [pricing page](https://vincevangen.com/pricing.md).

## Images

- `image:fast`: Drafts, variants, thumbnails. Model: `black-forest-labs/FLUX-1-schnell`
- `image:standard`: Cheap stills of people, scenes and illustration; exact pixel sizes; no reference images, no lettering. Model: `PrunaAI/p-image`
- `image:default`: The workhorse for finals; up to 4 reference images. Model: `black-forest-labs/FLUX-2-pro`
- `image:best`: Faces, hands and fine detail. Model: `black-forest-labs/FLUX-2-pro`
- `image:text`: Typography painted into the art. Model: `google/nano-banana-2-lite`
- `image:text-best`: Typography plus legible incidental signs in a busy scene. Model: `google/nano-banana-2`
- `image:premium`: Key art and covers. Model: `google/nano-banana-pro`

## Editing

- `edit:default`: Multi-reference edit and compose. Model: `black-forest-labs/FLUX-2-pro`
- `edit:precise`: Precise edits with a negative prompt; one image. Model: `Qwen/Qwen-Image-Edit`
- `edit:text`: Text edits, change only one thing. Model: `google/nano-banana-2-lite`
- `edit:text-best`: Text edits where every sign in the scene must stay legible. Model: `google/nano-banana-2`
- `edit:best`: Highest-fidelity edits. Model: `black-forest-labs/FLUX-2-pro`

## Video

- `video:fast`: Fast silent drafts. Model: `FastVideo/FastWan2.2-TI2V-5B-FullAttn-Diffusers`
- `video:default`: General text-to-video. Model: `Wan-AI/Wan2.6-T2V`
- `video:i2v`: Image-to-video with first/last frame. Model: `Wan-AI/Wan2.7-I2V`
- `video:best`: Higher quality, with audio. Model: `google/veo-3.1-fast`
- `video:premium`: Top quality. Model: `google/veo-3.1`

## Music

- `music:default`: Songs and instrumentals; lyrics, cover and repaint. Model: `ACE-Step/acestep-v15-xl-sft`

## Utilities

- `utility:remove_background`: Background removal. Model: `Bria/remove_background`
- `utility:expand`: Outpaint to a new aspect ratio. Model: `Bria/expand`

## Judging

- `vision:judge`: Default judge for pass/fail checks. Model: `Qwen/Qwen3-VL-30B-A3B-Instruct`
- `vision:judge_strict`: Stricter judge. Model: `Qwen/Qwen3-VL-235B-A22B-Instruct`

## Prompt enhancer

- `text:enhance`: Rewrites a brief into a model-ready prompt. Model: `deepseek-ai/DeepSeek-V4-Flash`

Aliases can move to a better model over time; every output records the model that made it in its provenance.