# NovAI

> NovAI is an AI API gateway providing OpenAI-compatible access to 85 Chinese frontier models — video (Seedance, Hailuo, Hunyuan, CogVideoX), image (Seedream, Hunyuan, CogView), 3D (Hunyuan 3D, Tripo), audio (MiniMax Speech & Music, Hunyuan / WAND ASR), and chat (DeepSeek, Qwen, GLM, Kimi, MiniMax, Doubao, Hunyuan). Official-direct endpoints, provider list pricing, zero platform fee. Credit card (Visa/Mastercard/UnionPay) and USDT (TRC20) accepted — instant top-up, no Chinese phone or Alipay needed.

## What NovAI does

NovAI aggregates Chinese AI models behind a single OpenAI-compatible API endpoint (base URL https://aiapi-pro.com/v1). Developers outside China get access without Chinese phone numbers, Alipay, or CNY billing. Every model is official-direct, priced at provider list rate with zero platform markup, launched day-one when upstream ships it.

## Key features

- Zero platform fee (vs OpenRouter's ~5.5% markup)
- OpenAI SDK compatible — change one line of code (base URL)
- 85 Chinese AI models through one API key
- $2 free credit on signup, no card required
- Free unlimited models: glm-4.6v-flash, glm-4.7-flash, glm-4.1v-thinking-flash (vision & chat), cogview-3-flash (image), cogvideox-flash (video)
- Video generation API (Seedance 2.5 with native audio): per-second billing, failure refunds
- Image generation API (Seedream 5.0, Hunyuan Image 3.0, Vidu, WAND Vega)
- Audio API (MiniMax Speech TTS, MiniMax Music song generation, Hunyuan / WAND speech-to-text): token-billed at $1.54/1M
- 3D generation API (Tencent Hunyuan 3D, Tripo / VAST): async text/image-to-3D, flat per-generation price from $0.185, refunded in full on failure
- Payment: credit card (Visa / Mastercard / UnionPay) and USDT (TRC20) — both LIVE now, instant top-up (credit confirmed in ~1 minute)
- Cost calculator, model finder, and live status page

## Video models

- doubao-seedance-2.5 (native synchronized audio, up to 15s): 480p $0.14/s, 720p $0.32/s
- doubao-seedance-2.0: 480p $0.067/s, 720p $0.163/s, 1080p $0.329/s, 4K $1.00/s
- doubao-seedance-2.0-fast: 480p $0.054/s, 720p $0.131/s, 1080p $0.264/s, 4K $0.802/s
- minimax-video-h3 (MiniMax Hailuo, T2V): $1.54/1M tokens (~$0.077/s, ~$0.385 per 5s clip)
- minimax-video-h3-max (MiniMax Hailuo, T2V top quality): $1.54/1M tokens (~$0.077/s)
- minimax-video-v2.3 (MiniMax Hailuo, T2V): $1.54/1M tokens (~$0.062/s, ~$0.308 per 5s clip)
- minimax-video-v2.3-fast (MiniMax Hailuo, image-to-video only): $1.54/1M tokens (~$0.042/s)
- hy-video-v1.5 (Tencent Hunyuan, T2V + I2V): $1.54/1M tokens (~$0.046/s, ~$0.231 per 5s clip)
- cogvideox-flash (Zhipu): FREE

## Image models

- doubao-seedream-5.0-pro: $0.034/image
- doubao-seedream-5.0: $0.027/image
- cogview-3-flash (Zhipu): FREE

TokenHub image models (via POST /v1/images/generations, billed by real upstream token usage; per-image figures are base-resolution references):
- hy-image-v3 (Tencent Hunyuan, sync): ~$0.031/image, up to 1024², 0-3 reference images
- vidu-image-q2 (Vidu / Shengshu, async auto-polled): from ~$0.029/image, 1080p-4K, 0-7 reference images
- seedream-image-v5.0-pro (ByteDance Seedream, sync): from ~$0.046/image, 1K-2K, layer decomposition, transparent background
- seedream-image-v5.0-lite (ByteDance Seedream, sync): ~$0.034/image, 2K-4K, group images, web-search tool
- wand-vega-image-lite (Tencent WAND Vega, async auto-polled): from ~$0.025/image, 1K-4K, 0-3 reference images
- wand-vega-image-flash (Tencent WAND Vega, async auto-polled): from ~$0.069/image, 1K-4K, 0-6 reference images
- wand-vega-image-pro (Tencent WAND Vega, async auto-polled): from ~$0.146/image, 1K-4K, layer output, 0-6 reference images

All image models return an OpenAI-compatible {created, data:[{url}]} response. Async upstream models (Vidu, WAND Vega) are polled to completion internally (up to ~180s) before the URL is returned. Charge = usage.total_tokens / 1000 × token_price.

## Free models (unlimited)

- glm-4.6v-flash (vision), glm-4.7-flash, glm-4.1v-thinking-flash (vision, thinking)
- cogview-3-flash (image), cogvideox-flash (video)
- Free image/video models are open to the public API with any API key - no top-up required.
- Free-tier API output carries a NovAI watermark; any completed top-up unlocks clean, no-watermark output.

## Chat models (USD per 1M tokens, input/output)

Moonshot Kimi:
- kimi-k3: $2.90 / $14.50, 1M context
- kimi-k2.8-preview: $0.93 / $3.89, 256K context
- kimi-k2.7-code: $0.67 / $3.40 (highspeed tier: $2.073 / $8.61)
- kimi-k2.6: $0.95 / $4.00
- kimi-k2.5: $0.60 / $3.00

Alibaba Qwen:
- qwen3.8-max: $1.84 / $5.515, 256K context
- qwen3.7-max: $1.16 / $3.475
- qwen3.6-max-preview: $1.03 / $6.16
- qwen3-max: $0.565 / $2.65
- qwen-plus: $0.185 / $0.53

DeepSeek:
- deepseek-v4-pro: $0.57 / $1.15, 1M context
- deepseek-v4-flash: $0.08 / $0.17, 1M context
- deepseek-v4.1-flash: $0.308 / $1.231, 128K context
- deepseek-v4-flash-vision (vision): $0.308 / $1.231

Zhipu GLM:
- glm-5.3: $1.25 / $4.00, 1M context
- glm-5.3-flash: $0.08 / $0.28, 1M context
- glm-5.2: $1.19 / $3.74, 128K context
- glm-5.1: $1.05 / $3.96
- glm-5: $0.60 / $1.92
- glm-5-turbo / glm-5v-turbo (vision): $1.08 / $3.692

MiniMax:
- minimax-m3: $0.30 / $1.20
- minimax-m2.7: $0.30 / $1.198
- minimax-m2.5: $0.27 / $1.08
- minimax-text-01: $0.14 / $1.10, 1M context

ByteDance Doubao:
- doubao-seed-2.0-pro: $0.40 / $2.00
- doubao-seed-2.0-code: $0.448 / $2.24
- doubao-seed-2.0-lite: $0.075 / $0.45

Tencent Hunyuan:
- hy3: $0.18 / $0.72
- hy3-preview: $0.127 / $0.423
- hy4-preview: $0.834 / $2.501, 1M context
- hunyuan-t1-vision (vision reasoning): $0.55 / $1.64
- hy-vision-2.0-instruct (vision): $1.154 / $2.692
- hunyuan-turbos-vision-video (image + video understanding): $0.462 / $1.385
- hy-mt2-pro (translation): $0.291 / $0.872
- hy-mt2-plus (translation): $0.072 / $0.287
- hy-mt2-lite (translation): $0.043 / $0.172
- hunyuan-role (roleplay): $0.465 / $1.861
- hy-role (roleplay): $0.369 / $1.477

Xiaomi MiMo:
- mimo-v2.5-pro: $0.427 / $0.855, 1M context

## Embedding models (USD per 1M input tokens)

Tencent Kinfra (via POST /v1/embeddings):
- kinfra-text-embedding-4b: $0.0923, 2560 dimensions (text)
- kinfra-text-embedding-0.6b: $0.0769, 1024 dimensions (text)
- kinfra-vl-embedding-8b: $0.0923, multimodal (text + image + video)
- kinfra-vl-embedding-2b: $0.0769, multimodal (text + image + video)

Text embedding models accept a string or list of strings as `input`. Multimodal (VL) models accept an object array such as [{"type":"text","text":"..."}] or [{"type":"image_url","image_url":"..."}]. Output vectors are free; only input tokens are billed.

## Audio models (token-billed at $1.54 / 1M tokens)

MiniMax Speech TTS (via POST /v1/audio/speech; returns an audio URL with response_format=url, or binary mp3/wav/opus bytes):
- minimax-speech-2.8-hd: latest flagship TTS, studio-grade voices + emotion control, ~$0.054 / 1K chars
- minimax-speech-2.8-turbo: fast low-latency TTS for realtime apps, ~$0.031 / 1K chars
- minimax-speech-2.6-hd: high-definition multilingual TTS, ~$0.054 / 1K chars
- minimax-speech-2.6-turbo: cost-optimized TTS at scale, ~$0.031 / 1K chars
- minimax-speech-02-hd: rich expressive timbres, ~$0.054 / 1K chars
- minimax-speech-02-turbo: economical speech generation, ~$0.031 / 1K chars

MiniMax Music (via POST /v1/audio/speech; full songs from lyrics + style prompt, returns audio URL):
- minimax-music-v3.0: latest song generation, vocal or instrumental, ~$0.154 / song
- minimax-music-v2.6: lyrics-to-song with configurable style, ~$0.154 / song

Speech-to-text / ASR (via POST /v1/audio/transcriptions; returns text with timestamps + SRT subtitle URL):
- hy-asr-3.0-preview (Tencent Hunyuan): accurate transcription, ~$0.003 / min audio
- wand-asr-v1 (Tencent WAND): multilingual + language detection, ~$0.007 / min audio

All audio is billed by actual upstream token usage (usage.total_tokens / 1000 × token_price = $1.54 / 1M, the same rate as image & video). Per-unit figures are references that scale with text length (TTS), track length (music) or audio duration (ASR).

## 3D models (flat per-generation price)

Tencent Hunyuan 3D & Tripo (VAST) (via POST /v1/three_d/generations + GET /v1/three_d/generations/{id}; async submit → poll → downloadable glb/obj/fbx/zip):
- hy-3d-express (Tencent Hunyuan, text/image-to-3D, obj zip): $0.277 / generation
- hy-3d-3.0 (Tencent Hunyuan, text/image-to-3D, obj zip + glb): $0.369 / generation
- hy-3d-3.1 (Tencent Hunyuan, text/image-to-3D, obj zip + glb): $0.369 / generation
- hy-3d-motion (Tencent Hunyuan, 3D motion, fbx): $0.185 / generation
- hy-3d-polygen-image (Tencent Hunyuan, image-to-3D required, glb + obj + fbx): $0.554 / generation
- tripo-3d-3.1 (Tripo / VAST, text/image-to-3D, glb): $0.323 / generation
- tripo-3d-p1 (Tripo / VAST, text/image-to-3D, glb): $0.323 / generation

Unlike image/video/audio, TokenHub bills 3D upstream by credits and returns no token usage on query, so NovAI charges a flat per-generation price (default tier). The full price is pre-deducted at submit and refunded in full if the job fails or is cancelled — you only pay for successful generations. Only the default generation mode is exposed (no PBR / high-poly / extra-format upcharges), so the flat price always matches the upstream cost. Submit returns {id, status:"queued"}; poll GET /v1/three_d/generations/{id} until status is "succeeded" to get model_url (primary, glb preferred) and outputs (every available format). Generation typically takes 1-8 minutes.

## API documentation

- Base URL: https://aiapi-pro.com/v1
- Compatible with OpenAI SDK (Python, JavaScript, Go, etc.)
- Supports: chat completions, streaming, function calling, Responses API, embeddings, image generation, audio (speech & transcription), video generation, 3D generation
- API key: Register at https://aiapi-pro.com/register ($2 free credit)

## Quick start

```python
from openai import OpenAI

client = OpenAI(
    api_key="sk-novai-your-key",
    base_url="https://aiapi-pro.com/v1"
)

response = client.chat.completions.create(
    model="deepseek-v4-flash",
    messages=[{"role": "user", "content": "Hello!"}]
)
```

## Featured guides (blog)
- https://aiapi-pro.com/blog/free-ai-image-video-generator-online-2026: Free AI Image & Video Generator Online: 20 Images + 5 Videos a Day, No Credit Card (2026)
- https://aiapi-pro.com/blog/codex-cli-novai-chinese-coding-models-setup-2026: Using OpenAI Codex CLI with NovAI: Run 45 Chinese Coding Models Through One Key (2026)

- https://aiapi-pro.com/blog/cheapest-chinese-llm-api-2026: Cheapest Chinese LLM APIs in 2026: Full Price Comparison
- https://aiapi-pro.com/blog/best-ai-coding-api-2026: Best AI Coding API in 2026: Kimi K2.7 Code, DeepSeek V4 & Qwen Compared
- https://aiapi-pro.com/blog/novai-vs-openrouter: NovAI vs OpenRouter: Pricing & Features Compared
- https://aiapi-pro.com/blog/novai-pricing-comparison-2026: NovAI Pricing Comparison 2026
- https://aiapi-pro.com/blog/chinese-llm-api-gateway-comparison-2026: Best Chinese LLM API Gateway 2026 Compared
- https://aiapi-pro.com/blog/doubao-api-outside-china: How to Access ByteDance Doubao API Outside China
- https://aiapi-pro.com/blog/doubao-api-pricing-breakdown: Doubao API Pricing: Real Cost Breakdown
- https://aiapi-pro.com/blog/doubao-seed-2-pro-benchmarks: Doubao Seed 2.0 Pro Benchmarks: Coding, Reasoning, Translation
- https://aiapi-pro.com/blog/kimi-k3-api-guide: Kimi K3 API Guide 2026
- https://aiapi-pro.com/blog/kimi-k2-7-code-api-guide-2026: Kimi K2.7 Code API Guide 2026
- https://aiapi-pro.com/blog/deepseek-api-guide-2026: DeepSeek API Guide 2026: Access Without a Chinese Phone
- https://aiapi-pro.com/blog/deepseek-v4-2m-token-context-window-explained: DeepSeek V4: 2M Token Context Window Explained
- https://aiapi-pro.com/blog/qwen38-max-api-guide: Qwen 3.8 Max API Guide 2026
- https://aiapi-pro.com/blog/qwen3-5-flash-cheapest-llm-api-2026: Qwen3.5 Flash Retired: The Cheapest Capable LLM APIs in 2026
- https://aiapi-pro.com/blog/glm-5-turbo-api-guide-2026: GLM-5 Turbo API Guide 2026
- https://aiapi-pro.com/blog/glm-5v-turbo-vision-api-multimodal-ai-at-budget-price: GLM-5V Turbo: Multimodal Vision API at Budget Price
- https://aiapi-pro.com/blog/minimax-m3-api-guide-2026: MiniMax M3 API Guide 2026
- https://aiapi-pro.com/blog/tencent-hunyuan-hy3-api-guide-2026: Tencent Hunyuan HY3 API Guide 2026
- https://aiapi-pro.com/blog/how-to-access-chinese-llm-apis-in-2026-full-guide: How to Access Chinese LLM APIs in 2026: Full Guide
- https://aiapi-pro.com/blog/responses-api-chinese-models: Responses API for Chinese Models
- https://aiapi-pro.com/blog/no-watermark-ai-video-api-test: No-Watermark AI Video API: Real Test Results
- https://aiapi-pro.com/blog/seedance-2-5-api-pricing: Seedance 2.5 API Pricing Guide
- https://aiapi-pro.com/blog/ai-video-generation-api-pricing-2026: AI Video Generation API Pricing 2026: Seedance vs Kling vs Hunyuan
- https://aiapi-pro.com/blog/pay-ai-api-usdt: Pay for AI APIs with USDT
- https://aiapi-pro.com/blog/paypal-ai-api-payment: PayPal Payment for AI APIs

Full blog index (180 articles): https://aiapi-pro.com/blog

## Links

- Website: https://aiapi-pro.com
- Model marketplace & pricing: https://aiapi-pro.com/model-marketplace.html
- NovAI vs OpenRouter: https://aiapi-pro.com/novai-vs-openrouter.html
- Model finder: https://aiapi-pro.com/model-finder.html
- Cost calculator: https://aiapi-pro.com/calculator.html
- Register: https://aiapi-pro.com/register
- API Docs: https://aiapi-pro.com/docs
- Status: https://aiapi-pro.com/status.html
- Video/Image playground: https://aiapi-pro.com/media.html
- Free image & video playground (no code, 20 images + 5 videos/day free): https://aiapi-pro.com/studio.html
- Blog: https://aiapi-pro.com/blog
- Full LLM fact sheet (all models, prices, guides): https://aiapi-pro.com/llms-full.txt

## Category landing pages

- Video generation API (Seedance, Hailuo H3, Hunyuan, free CogVideoX): https://aiapi-pro.com/video-generation-api.html
- Image generation API (Seedream, Hunyuan, Vega, free CogView): https://aiapi-pro.com/ai-image-generation-api.html
- Text-to-3D API (Hunyuan 3D, Tripo; glb/obj/fbx from $0.185): https://aiapi-pro.com/text-to-3d-api.html
- Text-to-speech & audio API (MiniMax Speech/Music, Hunyuan/WAND ASR): https://aiapi-pro.com/text-to-speech-api.html

### New this week (2026-09-16)

- [Text-to-3D API 2026: Hunyuan 3D + Tripo (glb/obj/fbx) from $0.185](https://aiapi-pro.com/blog/text-to-3d-api-hunyuan-tripo-2026)
- [MiniMax Speech TTS API vs ElevenLabs (2026): $0.054 vs $0.10 per 1K Characters](https://aiapi-pro.com/blog/minimax-speech-tts-api-vs-elevenlabs-2026)
- [Speech-to-Text API 2026: Hunyuan ASR $0.003/min vs OpenAI Whisper $0.006/min](https://aiapi-pro.com/blog/speech-to-text-api-chinese-asr-vs-whisper-2026)
- [Hailuo H3 Video API Guide 2026: MiniMax Flagship at ~$0.077/s](https://aiapi-pro.com/blog/hailuo-h3-video-api-guide-2026)
- [AI Music Generation API 2026: MiniMax Music vs Suno (Pay-Per-Song)](https://aiapi-pro.com/blog/ai-music-generation-api-minimax-vs-suno-2026)

### New this week (2026-09-13)

- [Use Chinese AI Model APIs Without a Chinese Phone Number: Complete Guide (2026)](https://aiapi-pro.com/blog/chinese-ai-api-without-chinese-phone-number-2026)
- [How to Use the GLM-5.3 API Without a Chinese Phone Number (2026)](https://aiapi-pro.com/blog/glm-5-3-api-without-chinese-phone-2026)
- [How to Use the Qwen3-Max API Without a Chinese Phone Number (2026)](https://aiapi-pro.com/blog/qwen3-max-api-without-chinese-phone-2026)
- [How to Use the Doubao Seed API Without a Chinese Phone Number (2026)](https://aiapi-pro.com/blog/doubao-seed-api-without-chinese-phone-2026)
- [How to Use the MiniMax M3 API Without a Chinese Phone Number (2026)](https://aiapi-pro.com/blog/minimax-m3-api-without-chinese-phone-2026)
- [Use Claude Code with Chinese Models via NovAI (claude-code-router Setup, 2026)](https://aiapi-pro.com/blog/claude-code-with-chinese-models-novai-2026)
- [Cline & Roo Code with Chinese Models: NovAI Setup Guide (2026)](https://aiapi-pro.com/blog/cline-roo-code-chinese-models-setup-2026)
- [Use Chinese Models in Cursor via NovAI: Base URL Override Guide (2026)](https://aiapi-pro.com/blog/cursor-chinese-models-novai-setup-2026)
- [Chatbox + Chinese Models: Connect GLM, Kimi & Qwen via NovAI (2026)](https://aiapi-pro.com/blog/chatbox-chinese-models-setup-2026)
- [Cherry Studio + Chinese Models: NovAI Provider Setup (2026)](https://aiapi-pro.com/blog/cherry-studio-chinese-models-setup-2026)

### New this week (2026-09-09)

- [How Much Does AI Video Generation Cost per Second? 2026 API Price Comparison](https://aiapi-pro.com/blog/ai-video-api-cost-per-second-2026)

- [DeepSeek API vs OpenAI (ChatGPT) API: Real Price Comparison 2026](https://aiapi-pro.com/blog/deepseek-vs-openai-api-pricing-2026)

- [Cheaper Than ChatGPT API: 7 OpenAI-Compatible Alternatives at a Fraction of the Cost (2026)](https://aiapi-pro.com/blog/cheaper-than-chatgpt-api-alternatives-2026)
- [GPT-5.5 vs Claude Opus 4.7 vs DeepSeek V4: Which API Should Developers Use?](https://aiapi-pro.com/blog/gpt-5-5-vs-claude-opus-4-7-vs-deepseek-v4-api)
- [DeepSeek V4 API Pricing: $0.08/1M In, 2M Context — Full Guide](https://aiapi-pro.com/blog/deepseek-v4-api-pricing-2m-context-guide)
- [Gemini 3.1 Pro API on a Budget: 5 Alternatives That Cost 90% Less](https://aiapi-pro.com/blog/gemini-3-1-pro-api-alternatives-budget-2026)
- [Claude Opus 4.7-Level API Access Cheaper: Pricing & Alternatives](https://aiapi-pro.com/blog/claude-opus-4-7-api-access-pricing-alternatives-2026)
- [Best LLM API Price-Performance Rankings — August 2026](https://aiapi-pro.com/blog/best-llm-api-price-performance-august-2026)
- [Qwen3.8-Max vs DeepSeek V4 Pro: Which One Should Power Your App?](https://aiapi-pro.com/blog/qwen-3-8-max-vs-deepseek-v4-pro-2026)
- [Kimi K3 API: Pricing, Benchmarks & Access Outside China (2026)](https://aiapi-pro.com/blog/kimi-k3-api-pricing-access-guide-2026)
- [Sora 2 vs Veo 3.1 vs Kling 3.0 vs Seedance: AI Video API Compared](https://aiapi-pro.com/blog/sora-2-vs-veo-3-1-vs-kling-vs-seedance-2026)
- [Free AI Video Generation API: CogVideoX Flash — No Credit Card](https://aiapi-pro.com/blog/free-ai-video-api-cogvideox-flash-2026)
- [Cheap AI Image APIs 2026: Seedream 5.0, Hunyuan & Free CogView](https://aiapi-pro.com/blog/cheap-ai-image-generation-api-2026-seedream-hunyuan)