WaveAPI 2.0 is live today at wave.qingbo.ai.
WaveAPI gives you one API key for text, image, video and audio models from many upstream providers, called the same way you call OpenAI. Version 2.0 rebuilds the model marketplace, the pricing page and the console, and adds a leaderboard ranked by real call volume. Your endpoint and your existing API keys stay the same.
What's new in 2.0
- Model marketplace: filter by modality, task, provider, capability and price; video and audio cards play examples in place.
- Run in the browser: sign in and run image, video and most audio models right on the model page, with live progress.
- Price before you run: see what a run will cost with your current settings; failed tasks are not charged.
- Pricing page: every model's price on one page, grouped by billing unit, with a monthly estimate.
- Rankings: ranked only by real calls on our platform, updated hourly.
- Console: usage and spend, call logs, task history, API keys and top-up, all rebuilt.
- English and Chinese: your first visit follows your browser language; switch any time.
- QingBo account: sign in with one QingBo account, which also works across QingBo's other products.
Model marketplace
In the model marketplace you can filter by the four modalities — text, image, video, audio — or by task, such as text-to-video, image editing or video extension. The sidebar narrows further by provider, capability and price range, and the search box looks across model names, tasks and providers at once.
Each card shows the provider, what the model does and its starting price. Video cards play a sample clip and audio cards let you listen, so you can get a feel for a model before opening it.
Run models on the model page
Every model has its own page listing the actions it supports, its parameters and how it is priced, followed by a code sample you can copy. Prices and code are visible without signing in.
Once signed in, you can run image, video and most audio models right on the page: set the parameters, submit, and watch the progress. For longer jobs such as video, you can close the page — the task keeps running on the gateway, and you can pick it up again in your task history.
Text models don't have an in-page chat. Their pages give you the code to call them: point your base_url at WaveAPI, set the model name, and it runs in your own app.
See the price before you run
Before you submit, the model page works out the cost for your current settings. Tasks billed per second, per image or per character get a quote; token-billed models get a cost ceiling for the run, and are settled on actual usage, never above that ceiling.
The amount is reserved when the task is submitted. If the upstream provider errors, times out or rejects the content, the reservation is released automatically and the task history marks it as released.
Pricing page
The pricing page puts every model's price on one page. Models that bill in the same unit sit together, cheapest first, with every price tier listed. A per-second price can't be compared with a per-image one, so they are grouped apart; within a group you can compare directly.
There is also a monthly estimator: enter your expected usage and get a rough monthly bill.
Rankings
Rankings use only real call records from the WaveAPI gateway — no estimates, no provider-reported figures.
How they are counted: text, image, video and audio are ranked separately; each ranking covers call volume over the last 30 days, and a model needs at least 100 calls to be listed; rankings update hourly and show each model's share and its change from the previous 30 days.
Console
The console is rebuilt from the ground up, in five parts:
- Usage and spend: see what you spent by day and by model. Failed calls are not counted.
- Call logs: every synchronous call, filterable by model.
- Task history: asynchronous image, video and audio tasks, showing the reserved amount while a task runs and the final charge once it settles. Filter by model or look up a task by its ID. Generated files are kept for 7 days.
- API keys: give each key its own spending limit and IP allowlist when you create it. The full key is shown only once, at creation.
- Top-up: pick a preset amount or enter your own. Your balance is in US dollars, never expires and has no minimum spend.
Also in this release
- The whole site is in English and Chinese. Your first visit follows your browser language; switch in the top-right corner.
- Sign in with a QingBo account, or create one from the same screen. The same account works across QingBo's other products.
- OpenAI SDK compatible: for text models, change the base_url and model name; image, video and audio all go through the
/v1/tasksendpoint.
If you already use WaveAPI
Your endpoint is still https://api.qingbo.ai/v1, and your existing API keys and balance carry over — no code changes needed. The console now uses QingBo account sign-in.
Get started
WaveAPI 2.0 is available to everyone from today.
If you run into a problem, check the help center first, and email us if the answer isn't there.