> ## Documentation Index > Fetch the complete documentation index at: https://docs.apiyi.com/llms.txt > Use this file to discover all available pages before exploring further. # Image & Video Generation Models > View supported image and video generation AI models with pricing and usage instructions. APIYI supports multiple image and video generation models. This page provides detailed model information, pricing, and usage instructions. For text and multimodal models, visit [Popular Models](/en/api-capabilities/model-info). ## 🎨 Image Generation Models | Model Name | Status | Features | Resolution/Specs | Price | | - | - | - | - | - | | [**Nano Banana Pro**](/en/api-capabilities/nano-banana-image-edit) πŸ”₯ | Best Seller | Knowledge understanding, great Chinese text, precise editing | 1K/2K/4K, supports original ratio editing | \$0.09/image | | [**Nano Banana 2**](/en/api-capabilities/nano-banana-2-image) πŸ”₯ | Hot, Fast | Same as Pro, supports usage-based billing | 0.5K-4K, new 1:8 and 8:1 ratios (long images) | \$0.055/image (usage-based \$0.025-0.07) | | [**Nano Banana Lite**](/en/api-capabilities/nano-banana-lite-image/overview) πŸ†• | New, Fastest & Cheapest | Google's fastest & cheapest, \~4s output, \~2.7x faster than NB2, 1K focused | 1K, 14 aspect ratios | \$0.025/image (usage-based \~\$0.018) | | [**gpt-image-2.5-flare**](/en/api-capabilities/gpt-image-2/overview) πŸ†• | New Sep 2026, Official | OpenAI GPT-Image 2.5 speed-first: higher quality than gpt-image-2 at up to 50% lower latency, precise size/quality control (adds `xhigh` / `max`), auto high-fidelity refs, mask inpainting | 1K/2K/**4K** (any valid size) | Token-billed, same price as gpt-image-2; \~15% off during top-up promos | | [**gpt-image-2.5-sunburst**](/en/api-capabilities/gpt-image-2/overview) πŸ†• | New Sep 2026, Official | OpenAI GPT-Image 2.5 quality- and editing-precision-first: stronger multi-turn edits and subject preservation, same parameters as flare | 1K/2K/**4K** (any valid size) | Token-billed, same price as gpt-image-2 | | [**gpt-image-2**](/en/api-capabilities/gpt-image-2/overview) | Official, previous gen | OpenAI official, precise size/quality control (up to `high`), auto high-fidelity refs, mask inpainting | 1K/2K/**4K** (any valid size) | Token-billed, standard rate; \~15% off during top-up promos | | [**gpt-image-2.5-all / 2-all**](/en/api-capabilities/gpt-image-2-all/overview) πŸ”₯ | Hot, Reverse | GPT reverse ChatGPT-web line, now on Images 2.5; both names share price and behavior, use `gpt-image-2.5-all` for new projects; strong text fidelity, native Chinese, faster output (\~30–60s), size in prompt | 1K-2K (prompt-controlled) | \$0.03/image (per-call) | | [**gpt-image-2.5-flare-vip / sunburst-vip**](/en/api-capabilities/gpt-image-2-vip/overview) πŸ†• | New Sep 2026, Reverse | Adobe-line reverse of GPT-Image 2.5: same call as -vip, `size` locks 30 presets incl. **4K**, plus all six `quality` tiers (incl. `xhigh` / `max`) and transparent background; alias `gpt-image-2.5-vip` | 1K/2K/**4K** (flat across 30 sizes) | \$0.03/image (per call) | | [**gpt-image-2-vip**](/en/api-capabilities/gpt-image-2-vip/overview) | Reverse | GPT reverse Adobe line (Firefly), identical call format to -all, `size` field locks dimensions, 30 explicit sizes incl. **4K**, \~90–150s | 1K/2K/**4K** (30 sizes, flat price) | \$0.03/image (per-call, no 4K surcharge) | | [**Nano Banana**](/en/api-capabilities/nano-banana-image) | Available | Fast, great consistency, e-commerce editing | Multiple sizes | \$0.02/image | | [**Seedream 5.0 Pro**](/en/api-capabilities/seedream-image/overview) πŸ†• | New, Pro tier | Best image quality and complex-instruction following in the family, interactive editing (coordinates / selection boxes / arrows), up to 10 reference images, png output; \~2 min per image, 5.0 Lite still the pick for everyday work | 1K/2K (total pixels ≀ 4.19M, no 3K/4K; no batch sequences or streaming) | \$0.12/call | | [**Seedream 5.0 Flash**](/en/api-capabilities/seedream-image/overview) πŸ†• | New Sep 2026, cheapest in the family | BytePlus fast tier, \~15–20s per image, accurate Chinese typography, reference images free; no batch sequences or streaming, watermarked by default (pass `watermark: false` to turn off) | Multiple sizes | \$0.018/image | | [**Seedream 5.0 Lite**](/en/api-capabilities/seedream-image) | Available | Price advantage, fast, URL output | 2K/3K | \$0.035/image | | [**Seedream 4.5**](/en/api-capabilities/seedream-image) | Available | Price advantage, fast, URL output | 2K/4K | \$0.04/image | | [**Seedream 4.0**](/en/api-capabilities/seedream-image) | Available | Price advantage, fast, URL output | 2K | \$0.035/image | | [**GPT Image 1.5**](/en/news/gpt-image-1-5-launch) πŸ”₯ | Official | Precise editing, 4x speed boost, enhanced text rendering | Low/Med/High quality | Usage-based | | [**GPT Image 1**](/en/api-capabilities/gpt-image-1) | Official | Precise editing | Multiple sizes | Usage-based | | **GPT Image 1-Mini** | Official | Alternative to sora\_image | Multiple sizes | Usage-based | | [**Grok Imagine 2**](/en/api-capabilities/grok-imagine-image/overview) πŸ†• | New, πŸ”’ by application | xAI's second-gen image model, standard `grok-imagine-image` / high-quality `grok-imagine-image-quality`; 5 aspect ratios, up to 10 images per call, true reference editing; requires the dedicated `Grok_imagine` group | 1K/2K (same price) | \$0.02 / \$0.045 per image | | [**flux-2-max**](/en/api-capabilities/flux/overview) πŸ†• | Latest gen (FLUX.2) | FLUX.2 flagship, high-quality output | Multiple sizes | \$0.07/call | | [**flux-2-pro**](/en/api-capabilities/flux/overview) πŸ†• | Latest gen (FLUX.2) | FLUX.2 pro, balanced quality/price | Multiple sizes | \$0.03/call | | [**flux-2-flex**](/en/api-capabilities/flux/overview) πŸ†• | Latest gen (FLUX.2) | FLUX.2 flex, tunable parameters | Multiple sizes | \$0.06/call | | [**flux-2-klein-4b / 9b**](/en/api-capabilities/flux/overview) | Latest gen (FLUX.2) | FLUX.2 lightweight, sub-second output, up to 4 reference images | Multiple sizes | \$0.01/call | | [**Flux Kontext Pro**](/en/api-capabilities/flux-image-generation) | Available | Image editing | Multiple sizes | See docs | | [**Flux Kontext Max**](/en/api-capabilities/flux-image-generation) | Available | High-quality image editing | Multiple sizes | See docs | Model prices are aligned with the official sites and may change with them; the tables above are for reference only, and the **Model Pricing** tab in the top navigation is authoritative: [Model Pricing](/en/models/index). **Nano Banana Pro Special**: All resolutions from 1K-4K at a uniform price of \$0.09/image. Official 4K price is \$0.24/image β€” \~38% of official pricing! NanoBananaEnterprise enterprise HA channel also available at 1.4x rate (\$0.126/image). [View details](/en/news/nano-banana-pro-launch) **Image Generation Testing Tools** * China access: image.apiyi.com * Global access: imagen.apiyi.com Detailed Documentation: * [Nano Banana Pro Docs](/en/api-capabilities/nano-banana-image-edit) - Best on platform, knowledge understanding + precise editing * [Nano Banana 2 Docs](/en/api-capabilities/nano-banana-2-image) - Usage-based billing, new ultra-wide ratios * [gpt-image-2.5-all / 2-all Docs](/en/api-capabilities/gpt-image-2-all/overview) - GPT reverse ChatGPT-web line (now Images 2.5), \$0.03/image, \~30–60s faster output * [gpt-image-2-vip Docs](/en/api-capabilities/gpt-image-2-vip/overview) - GPT reverse Adobe line (Firefly), \$0.03/image, 30 sizes incl. 4K, \~90–150s * [GPT-Image-2.5 / 2 Docs](/en/api-capabilities/gpt-image-2/overview) - OpenAI official trio (2.5-flare / 2.5-sunburst / 2), native 4K, precise size/quality control * [βš–οΈ Official vs Reverse Comparison](/en/api-capabilities/gpt-image-2/vs-gpt-image-2-all) - gpt-image-2 vs reverse siblings -all / -vip selection guide * [Nano Banana Docs](/en/api-capabilities/nano-banana-image) - Fast, great consistency * [Nano Banana Lite Docs](/en/api-capabilities/nano-banana-lite-image/overview) - Fastest & cheapest, \~4s output, \$0.025/image per-call * [Seedream Docs](/en/api-capabilities/seedream-image) - Price advantage, fast; includes 5.0 Flash (\$0.018/image, \~15–20s) and the 5.0 Pro tier (\$0.12/call, \~2 min per image) * [Grok Imagine 2 Docs](/en/api-capabilities/grok-imagine-image/overview) - xAI's second-gen image model, \$0.02 / \$0.045 per image, dedicated group by application * [GPT Image 1.5 Docs](/en/news/gpt-image-1-5-launch) - 4x speed boost, precise editing * [GPT Image 1 Docs](/en/api-capabilities/gpt-image-1) - Official image generation * [FLUX.2 Docs](/en/api-capabilities/flux/overview) - Full klein / pro / flex / max lineup, text-to-image and multi-image fusion * [Flux Kontext Docs](/en/api-capabilities/flux-image-generation) - Image editing ## 🎬 Video Generation Models | Model Name | Status | Features | Duration | Price | | - | - | - | - | - | | [**Seedance 2.5 / 2.0 Series**](/en/api-capabilities/seedance2/overview) πŸ”₯ | Live, Hot | ByteDance, official Volcengine China resources; **2.5** plus 2.0 standard / `fast` / `mini` in parallel β€” text-to-video, first/last frame, multimodal references (image + video + audio), video editing and extension, **synced audio by default**, private asset library included free | 4–30s on 2.5, 4–15s on the 2.0 family (480p/720p/1080p; 1080p on 2.5 and standard only) | Token-billed (area Γ— duration); measured 720p/5s: \$0.45 (mini) / \$0.73 (fast) / \$0.91 (standard) / \$1.35 (2.5) | | [**Wan2.7 Video**](/en/api-capabilities/wan/overview) πŸ†• | New, Recommended | Alibaba Wanxiang, text/image/reference/video-edit generation, i2v supports audio-driven | Per-second (max 12s) | \$0.084-0.14/sec (98% of official) | | [**Wan2.6 Video**](/en/api-capabilities/wan/historical-versions) | Available | Shares endpoint with Wan2.7, includes `r2v-flash` low-latency tier | Per-second | See console | | [**VEO 3.1 Official**](/en/api-capabilities/veo-3-1-official/overview) πŸ”₯ | Live, Hot (reverse alternative) | Passthrough to Google AI Studio official endpoint, audio-video sync, real people, callable from default group | 4/6/8s | Per-call \$0.3 / \$1.2 (720p/1080p/4k same price) | | **Veo 3.1 Reverse** | ⏸️ Paused (Google risk control) | Google Flow reverse; use **VEO 3.1 Official** during the pause | Fixed 8s | Unavailable | | **Sora 2 Official** | ⏸️ Discontinued | OpenAI official relay, professional creation, high stability, supports sora-2-pro, no "character" reference | 4/8/12s | Unavailable | | [**HappyHorse 1.1**](/en/api-capabilities/happyhorse/overview) πŸ†• | New | Alibaba, multi-reference subject consistency (up to 9 images), shares group with Wan | Per-second (max 12s) | \$0.126-0.224/sec (98% of official) | | [**MiniMax H3**](/en/api-capabilities/minimax-h3/overview) πŸ†• | New Sep 2026, self-hosted open weights | Hailuo 3.0 on our own deployment of the open weights; one endpoint for text, first/last-frame, and mixed image/video/audio references, stereo sound, 768P; reference media free, default group | 4–15s | \$0.03/sec | | [**Oxygen**](/en/api-capabilities/oxygen/overview) πŸ†• | New Oct 2026, built for volume | AZ8's in-house all-round video model, OpenAI Videos-compatible: text, first-frame, first-and-last-frame, and reference image + video + audio, built-in audio, 320p–768p at one price; default group | 4–15s | \$0.02/sec | | ~~**Sora 2 Reverse**~~ | ❌ Discontinued | Formerly e-commerce / anime scenarios, use Sora 2 Official instead | β€” | β€” | Model prices are aligned with the official sites and may change with them; the tables above are for reference only, and the **Model Pricing** tab in the top navigation is authoritative: [Model Pricing](/en/models/index). **Video Model Key Features**: * **Seedance 2.5 / 2.0 Series (ByteDance)**: `doubao-seedance-2-5-260628` (2.5 β€” up to 30 s, 30 reference images, mov output) / `doubao-seedance-2-0-260128` (standard) / `-fast-260128` (fast) / `-mini-260615` (mini) β€” prices and speeds differ (mini \< fast \< standard \< 2.5, with 2.5 at about 1.5Γ— standard); mini and fast top out at 720p. **All four models run on the `SeeDance2` group** (0.18x), so one token covers them all. The token's billing mode must be "usage-first" or "usage-based". `SD2Mini` (0.10x) and `SD2Fast` (0.15x) are limited-time discount groups β€” **44.4% off for mini, 16.7% off for fast, through 2026-10-07 23:59 (UTC+8)**; after that the rate returns to 0.18x and the groups stay available * **VEO 3.1 Official**: Passthrough to Google AI Studio official endpoint, industry-leading audio-video sync, supports real people, callable with default group + per-call/usage-first tokens, the recommended alternative while Veo 3.1 Reverse is paused * **Veo 3.1 Reverse**: **Paused** due to Google risk control, recovery time TBA, use VEO 3.1 Official in the meantime * **Sora 2 Official**: Discontinued, currently unavailable * **Wan2.7 / Wan2.6 (Alibaba Wanxiang)**: Default price \~98% of official, both series share the `Wan&HappyHorse` group with HappyHorse, one token for all; Wan2.6 and Wan2.7 share the same endpoint and schema, just change the `model` name to migrate * **HappyHorse 1.1 (Alibaba)**: Excels at multi-reference subject consistency, shares the `Wan&HappyHorse` group and endpoint with Wan * **MiniMax H3 (self-hosted open weights)**: `MiniMax-H3` on `POST /hailuo/v2/video_generation`, up to 9 reference images + 3 reference videos + 3 reference audio clips, billed by requested duration; 2K is not supported yet and the `Idempotency-Key` header has no effect, so handle idempotency in your application * **Oxygen (AZ8)**: `oxygen-1.0` on `POST /v1/videos`; always pass `size` (portrait is the default otherwise), and put first-and-last frames, reference media, 320p, and other advanced options inside the JSON envelope in `input_reference`; failed tasks are refunded in full automatically * AI video tool (online testing): icover.ai **Upcoming Video Models**: * Kling 3.0 Stay tuned! Follow us for the latest launch notifications. ## πŸ’° Pricing Information * **Pay-as-you-go**: Charged based on actual usage, no minimum charge * **Balance validity**: Valid for 365 days from the top-up date, automatically reset on the next top-up (see [Recharge Promotions](/en/faq/recharge-promotions)) * Visit [APIYI Console Pricing Page](https://www.apiyi.com/account/pricing) for latest pricing on all models This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.