Models & Options

The catalog endpoint GET /v1/models returns this same data programmatically. For a single model with its full options matrix call GET /v1/models/{model_id}.

Orientation cheatsheet

You want	`ratio`
Vertical / portrait (TikTok, Reels, Shorts, Stories)	`9:16`
Horizontal / landscape (YouTube, web hero)	`16:9`
Square (Instagram feed)	`1:1`
Ultrawide / cinematic	`21:9`

The MCP server and CLI accept natural-language aliases (vertical, portrait, tiktok, horizontal, landscape, square, ig, ultrawide, cinematic). The HTTP API itself requires the W:H string.

Some video models (wan-2.6, hailuo-2.3, grok-imagine-1.5, wan-2.6-v2v, kling-2.6-mc, kling-3.0-mc) inherit the aspect ratio from their input image or video and reject the ratio parameter. Omit ratio for those models. minimax-h3 inherits only on the image-to-video path: send ratio for text-to-video, but omit it when input_image_url is set — the output matches the input image.

Image models

Model	Qualities	Ratios	Max ref images	Max outputs / run
`nano-banana-pro`	`1K`, `2K`, `4K`	1:1, 2:3, 3:2, 3:4, 4:3, 4:5, 5:4, 9:16, 16:9, 21:9	8	4
`nano-banana-2`	`1K`, `2K`, `4K`	1:1, 1:4, 1:8, 2:3, 3:2, 3:4, 4:1, 4:3, 4:5, 5:4, 8:1, 9:16, 16:9, 21:9	14	4
`nano-banana-2-lite`	`default` (fixed 1K)	1:1, 1:4, 1:8, 2:3, 3:2, 3:4, 4:1, 4:3, 4:5, 5:4, 8:1, 9:16, 16:9, 21:9	10	4
`flux-2-pro`	`1K`, `2K`	1:1, 4:3, 3:4, 16:9, 9:16, 3:2, 2:3	8	4
`flux-2`	`1K`, `2K`	1:1, 4:3, 3:4, 16:9, 9:16, 3:2, 2:3	8	4
`seedream-4-5`	`default`	1:1, 4:3, 3:4, 16:9, 9:16, 2:3, 3:2, 21:9	14	4
`wan-2-7-pro`	`default`	1:1, 16:9, 4:3, 21:9, 3:4, 9:16, 8:1, 1:8	9	4
`wan-2-7`	`default`	1:1, 16:9, 4:3, 21:9, 3:4, 9:16, 8:1, 1:8	9	4
`ideogram-v3`	`TURBO`, `BALANCED`, `QUALITY`	1:1, 4:3, 3:4, 16:9, 9:16	1	4
`grok-imagine`	`STANDARD`, `QUALITY`	1:1, 2:3, 3:2, 16:9, 9:16	1	2–6 per run
`qwen2`	`default`	1:1, 3:4, 4:3, 9:16, 16:9	1	4
`seedream-5-lite`	`basic`, `high`	1:1, 4:3, 3:4, 16:9, 9:16, 2:3, 3:2, 21:9	14	4
`gpt-image-2`	`1K`, `2K`, `4K`	1:1, 2:3, 3:2, 3:4, 4:3, 16:9, 9:16, 21:9	16	4
`gpt-image-1-5`	`medium`, `high`	1:1, 2:3, 3:2	16	4
`z-image`	`default`	1:1, 4:3, 3:4, 16:9, 9:16	0 (text-to-image only)	4

Image credits per quality

Model	Credits per image
`nano-banana-pro`	1K: 12 · 2K: 12 · 4K: 15
`nano-banana-2`	1K: 5 · 2K: 8 · 4K: 12
`nano-banana-2-lite`	default: 3
`flux-2-pro`	1K: 4 · 2K: 5
`flux-2`	1K: 9 · 2K: 15
`seedream-4-5`	default: 5
`wan-2-7-pro`	default: 8
`wan-2-7`	default: 3
`ideogram-v3`	TURBO: 3 · BALANCED: 5 · QUALITY: 7
`grok-imagine`	STANDARD: 3 · QUALITY: 4
`qwen2`	default: 4
`seedream-5-lite`	basic: 3 · high: 4
`gpt-image-2`	1K: 3 · 2K: 5 · 4K: 8
`gpt-image-1-5`	medium: 2 · high: 11
`z-image`	default: 1

Video models

The audio_mode column tells you whether you should send audio_enabled:

toggle — pass true or false.
forced-on — the model always returns audio; sending audio_enabled=false is rejected.
forced-off — the model never returns audio; sending audio_enabled=true is rejected.
n/a — the field is ignored entirely.

Model	Mode	Ratios	Resolutions	Durations (s)	Tiers	Audio	Input modes
`veo-3.1`	`generate`, `edit` (extend)	9:16, 16:9	720p, 1080p, 4K	8	lite, fast, quality	forced-on	text, first-frame image, first+last frame
`kling-2.6`	`generate`	9:16, 16:9, 1:1	1080p	5, 10	—	toggle	text, first-frame image
`kling-3.0`	`generate`	9:16, 16:9, 1:1	720p, 1080p	3, 5, 7, 10, 15	—	toggle	text, first-frame image
`seedance-1.5-pro`	`generate`	1:1, 4:3, 3:4, 16:9, 9:16, 21:9	480p, 720p, 1080p	4, 8, 12	—	toggle	text, first-frame image
`seedance-2.0`	`generate`	1:1, 4:3, 3:4, 16:9, 9:16, 21:9, adaptive	480p, 720p, 1080p*	4, 8, 12	fast, quality	n/a	text, first-frame image
`seedance-2.5`	`generate`	1:1, 4:3, 3:4, 16:9, 9:16, 21:9, adaptive	480p, 720p	4, 5, 8, 10, 15, 20, 30	—	toggle	text, first-frame image, first+last frame, 1–4 reference images
`wan-2.6`	`generate`	inherit	720p, 1080p	5, 10, 15	—	forced-on	first-frame image (required)
`wan-2.7`	`generate`	16:9, 9:16, 1:1, 4:3, 3:4	720p, 1080p	5, 8	—	forced-off	text, first-frame image, 1–4 reference images
`hailuo-2.3`	`generate`	inherit	768p, 1080p	6, 10†	standard, pro	forced-off	first-frame image (required)
`minimax-h3`	`generate`	21:9, 16:9, 4:3, 1:1, 3:4, 9:16‡	768p, 2K	4–15	—	forced-on	text, first-frame image (optional)
`grok-imagine`	`generate`	2:3, 3:2, 1:1, 16:9, 9:16	480p, 720p	6, 10, 15, 20, 30	fun, normal, spicy	forced-off	text, first-frame image
`grok-imagine-1.5`	`generate`	inherit	480p, 720p	6, 8, 10, 15	—	forced-off	first-frame image (required)
`happyhorse-1.1`	`generate`	16:9, 9:16, 1:1, 4:3, 3:4, 4:5, 5:4, 9:21, 21:9	720p, 1080p	5, 8, 10, 15	—	n/a	text, first-frame image, 1–4 reference images
`kling-3.0-turbo`	`generate`	1:1, 9:16, 16:9	720p, 1080p	5, 8, 10, 15	—	n/a	text, first-frame image
`gemini-omni`	`generate`	16:9, 9:16	720p, 1080p, 4K	4, 6, 8, 10	—	n/a	text, up to 7 reference images
`sora-2`	`generate`	9:16, 16:9	720p	4, 8, 12, 16, 20	—	forced-on	text, first-frame image
`sora-2-pro`	`generate`	9:16, 16:9	720p, 1024x1792, 1080x1920	4, 8, 12, 16, 20	—	forced-on	text, first-frame image
`wan-2.6-v2v`	`edit`	inherit	720p, 1080p	5, 10	—	forced-off	input_video_url (required)
`wan-2.7-videoedit`	`edit`	16:9, 9:16, 1:1, 4:3, 3:4	720p, 1080p	5, 8, 10	—	forced-off	input_video_url (required)
`seedance-2.5-v2v`	`edit`	1:1, 4:3, 3:4, 16:9, 9:16, 21:9, adaptive	480p, 720p	4, 5, 8, 10, 15, 20, 30	—	toggle	input_video_url (required), optional reference images
`minimax-h3-v2v`	`edit`	21:9, 16:9, 4:3, 1:1, 3:4, 9:16, adaptive	768p, 2K	4–15	—	forced-on	input_video_url (required)
`kling-2.6-mc`	`motion`	inherit	inherit	inherit	—	forced-off	input_image_url + input_video_url
`kling-3.0-mc`	`motion`	inherit	inherit	inherit	—	forced-off	input_image_url + input_video_url

* seedance-2.0 with tier=fast does not support 1080p.

seedance-2.5-v2v is billed on your reference video’s length as well as the output’s. The upstream model charges per second of input + output, so a short reference costs less than a long one:credits = roundUpTo10( rate × ( min(ceil(reference_seconds), 30) + duration_seconds ) ) with rate = 8.5 at 480p and 19 at 720p.A 10s reference with a 10s output at 480p is 170 credits; the same output with a 30s reference is 340. The /v1/pricing rows list the maximum (a 30s reference) — you are charged the measured amount, never more. If the reference’s duration cannot be read, the maximum applies.

seedance-2.5 and seedance-2.5-v2v top out at 720p — there is no 1080p or 4K option on that endpoint. They are the only models that generate a full 30 seconds in a single pass.† hailuo-2.3 with resolution=1080p is locked to duration_seconds="6" (no 10s option).‡ minimax-h3 requires ratio for text-to-video (there is no adaptive), but ignores it when input_image_url is set — the output matches the input image, so omit ratio for image-to-video. Both H3 ids always generate native stereo audio (forced-on, no audio flag).

minimax-h3-v2v is billed on your reference video’s length as well as the output’s. The upstream model charges per second of input + output — 8 credits/s at 768p, 13 credits/s at 2K, rounded up to the nearest 10. The /v1/pricing rows list the maximum (a 15s reference) — shorter reference videos are charged on their measured duration, never more.Reference videos must be MP4/MOV (H.264/H.265), ≤ 50 MB, 2–15s per clip, and ≤ 15s total.

veo-3.1 with tier=lite cannot be extended (the extend endpoint will reject it). Extensions cost 60 credits each.grok-imagine with tier=spicy rejects user-uploaded reference images.veo-3.1 “first-last” input mode sends both a start and end frame image — this is currently accessible via the unified video studio dispatcher; pass two URLs through the storyboard flow (the public API surface accepts a single input_image_url for the start frame). Call GET /v1/models/veo-3.1 for the current accepted combinations.

Video credits (selected combinations)

Call GET /v1/pricing?type=video for the authoritative matrix. Examples:

Model	tier	resolution	duration	audio	credits
`veo-3.1`	fast	1080p	8	true	60
`veo-3.1`	quality	4K	8	true	180
`kling-2.6`	—	1080p	5	true	30
`kling-3.0`	—	1080p	10	true	60
`seedance-1.5-pro`	—	720p	4	false	15
`seedance-2.0`	quality	1080p	8	n/a	40
`seedance-2.5`	—	480p	4	true	50
`seedance-2.5`	—	720p	30	true	960
`seedance-2.5-v2v`	—	480p	10	true	340 (30s reference) · 170 (10s reference)
`wan-2.6`	—	1080p	5	true	25
`wan-2.7`	—	1080p	5	false	18
`hailuo-2.3`	pro	1080p	6	false	28
`minimax-h3`	—	768p	6	true	50
`minimax-h3`	—	2K	6	true	80
`minimax-h3`	—	768p	15	true	120
`minimax-h3`	—	2K	15	true	195
`minimax-h3-v2v`	—	768p	6	true	170 (15s-reference ceiling)
`minimax-h3-v2v`	—	2K	6	true	280 (15s-reference ceiling)
`grok-imagine`	normal	720p	10	false	20
`grok-imagine-1.5`	—	720p	6	false	13
`happyhorse-1.1`	—	720p	5	n/a	50
`kling-3.0-turbo`	—	720p	5	n/a	45
`gemini-omni`	—	1080p	8	n/a	75
`sora-2`	—	720p	8	true	35
`sora-2-pro`	—	1080x1920	8	true	80

Voice (TTS / STS) model

Field	Value
`model_id`	`elevenlabs-multilingual-v2`
Modes	`tts` (text-to-speech), `sts` (speech-to-speech)
Voice sources	Library voices (`GET /v1/voices`) and your custom clones (`GET /v1/custom-voices`)
Output format	`mp3`
TTS max text	8192 characters
TTS duration estimate	`chars / 12` seconds
STS audio formats	`mp3`, `wav`, `m4a`, `ogg`, `webm` (≤ 25 MB)
Pricing	2 credits per 10 seconds, rounded up. Minimum 2 credits.

Talking-avatar model

Field	Value
`model_id`	`adskull-talking-avatar`
Modes	`script` (TTS + lip-sync), `audio` (lip-sync to an existing file)
Avatar image	png/jpg/jpeg/webp, ≤ 25 MB, single front-facing portrait
Script min/max	100 chars / 8192 chars
Script duration estimate	`chars / 10` seconds
Audio formats	`mp3`, `wav`, `m4a` (≤ 25 MB)
Product placement	`product_placement=true` + `product_image_url` adds 5 credits
Background overlay	Optional `background_url`
Pricing	3 credits per second, minimum 5 seconds

Start Here

Features

Core Concepts

Orientation cheatsheet

Image models

Image credits per quality

Video models

Video credits (selected combinations)

Voice (TTS / STS) model

Talking-avatar model

​Orientation cheatsheet

​Image models

​Image credits per quality

​Video models

​Video credits (selected combinations)

​Voice (TTS / STS) model

​Talking-avatar model

Orientation cheatsheet

Image models

Image credits per quality

Video models

Video credits (selected combinations)

Voice (TTS / STS) model

Talking-avatar model