Reference
Models
The model catalog available through the studio, MCP, and CLI.
Image
| ID | Strength | Base credits |
|---|---|---|
| soul | Fashion-grade realism, 89 aesthetic presets | 4/image |
| ideogram-4 | Typography, posters, packaging, and precise layouts | 2/image |
| qwen-image-3 | Detailed text and instruction-led editing with up to three references | 4/image |
| flux-pro | Precise prompt following, crisp detail | 8/image |
| gpt-image-2 | Photorealism, typography, and complex instructions | 4/image |
| seedream-5-pro | Flagship visual reasoning and multilingual text | 4/image |
| seedream-5-lite | Fast high-resolution creative generation | 3/image |
| nano-banana-2 | Fast reasoning, text, and reference consistency | 5/image |
| nano-banana-pro | Maximum compositional reasoning up to 4K | 8/image |
| nano-banana-2-lite | Sub-two-second generation at fixed 1K | 3/image |
| grok-imagine-2 | Versatile photoreal and stylized generation | 4/image |
| flux-2-pro | Production-ready detail and consistency | 3/image |
| flux-2-flex | Fine-grained control over precision and speed | 6/image |
| recraft-v4 | Design-led imagery and brand-ready composition | 3/image |
| recraft-v4-pro | Premium design and photoreal production assets | 13/image |
| z-image | Fast, economical generation for high-volume work | 2/image |
| wan-2.2-image | High-fidelity cinematic still generation | 1/image |
| kling-o1-image | Precise multi-reference composition | 2/image |
| flux-2-max | Highest-quality reference editing | 15/image |
| flux-kontext-max | Premium instruction-led image editing | 5/image |
Video
Video pricing depends on duration, resolution, quality, and engine. Call estimate_generation_cost before submitting. Reference-video quotes inspect your owned uploaded assets.
- foxcut-dop — camera-directed motion
- minimax-h3-max — ultra-fast 480p/768p native-audio video with first/last frames or image/video/audio references
- seedance-2.0 and seedance-2.5 — multi-shot continuity
- kling-3.0 and kling-2.5 — photoreal motion and frame control
- sora-2 and veo-3.1 — high-end world simulation
- veo-3.1-fast — 4/6/8-second native-audio video at 720p or 1080p
- wan-3.0 — native-audio video, 2–30 seconds, first/last frames
- minimax-hailuo and wan-2.5 — fast or value-oriented motion
Standalone speech audio
generate_audio creates a private WAV without a portrait. Use list_voices for a voice ID and list_models with media_type=audio for Fish model IDs. speech_model accepts s2.1-pro-free or s2.1-pro; omission uses the server default. foxcut-speak is the FoxCut product ID used for credit estimates, not the underlying Fish version.
Talking avatar
- foxcut-speak — native speech and lip sync
- latentsync — fast audio-driven synchronization
- kling-avatar — expressive full-scene performance