A web app for WaveSpeed batch generation
Updated October 28, 2026 · By Personifex
Personifex is a web app that runs WaveSpeed's image and video models in bulk, on your own WaveSpeed API key. You queue many different prompts, see what the whole batch costs before you press Start, and up to 20, 50 or 300 run at once depending on the plan, once your WaveSpeed account has had its first top-up. Failed images retry on their own, and the results land in a review pass where you keep the best.
Personifex is an independent product, not made by WaveSpeed. It is a browser UI that calls WaveSpeed's API with the key you paste in, and WaveSpeed bills that account at its own prices.
What WaveSpeed itself offers
WaveSpeed hosts the models and sells the compute. Its docs list several ways to run them:
- An Image Generator and a Video Generator in the browser, with curated models and no code. The Image Generator can make several variations of one prompt in a run.
- A REST API, with Python and JavaScript SDKs, for developers.
- A desktop app for Windows, macOS and Linux, and integrations for ComfyUI and n8n.
Read from wavespeed.ai on October 28, 2026
If you write code, the API can already run as many tasks at once as your WaveSpeed account level allows. A batch layer is for the rest of the job without code: many different prompts at once, the price up front, retries, and a fast way through hundreds of results.
What a batch layer adds
- A queue of different prompts for each persona (an AI person with its own reference photos) or product. Write them, paste them, import a text file, or have the Prompt maker draft them on your own OpenRouter key.
- Your reference photos go with every prompt, so the model works from the same face or product each time.
- The whole batch's price on the Start button before anything runs.
- As many at once as your plan allows. Start takes as many waiting prompts as fit and leaves the rest queued for the next batch.
- Retries: a failed image runs again on its own, up to 4 tries in all.
- A review pass: every result full size, rated with one key, the keepers sent to the Gallery.
- On Creator and Agency, a calendar and a posting queue for the team.
Which models it runs
It runs 123 image models and 187 video models, all hosted by WaveSpeed and all on your one key. Among them:
| Family | Models | What they do |
|---|---|---|
| Google Nano-Banana | 14 | image edit, text-to-image |
| ByteDance Seedream | 10 | image edit, text-to-image |
| Kuaishou Kling | 27 | image edit, text-to-image, image-to-video, reference-to-video, text-to-video, change a video, effects and templates, motion control, talking avatar |
In the app every model goes by its WaveSpeed endpoint name, such as google/nano-banana-pro/edit, bytedance/seedream-v5.0-pro/edit or kwaivgi/kling-v3.0-pro/image-to-video. Each has a page with its full price grid on the model list, and the cheapest model for each job ranks them for common jobs.
How many run at once
The plan sets how many generations run at the same time:
| Plan | A month | At once |
|---|---|---|
| Solo | $29 | 20 |
| Creator | $79 | 50 |
| Agency | $179 | 300 |
The limit counts every image still to make across the workspace, so teammates' runs share it. A new WaveSpeed account runs only two at a time until its first top-up, of any size, so a batch crawls until then: setting up a WaveSpeed key covers that step.
See the price, then let it run
Before anything runs, the Start button shows what the whole batch will cost, and the line above it shows the price of each image on the model and resolution you picked. It is WaveSpeed's list price, or its sale price while WaveSpeed runs a sale, and WaveSpeed bills it to your own account. Personifex adds nothing per image.
At volume some calls fail. The Passes setting retries a failed image on its own, up to 4 tries in all, and an extra pass costs nothing unless something fails.
Every result then lands in Review, full size and one at a time, and one key rates it: 1 for Trash, 2 for Good, 3 for Great, 4 for Needs edit. Picking the best of a big batch covers the method, and generating images in bulk walks through a whole run.
What it does not do
- It needs your own WaveSpeed account with credit on it. There is no compute to buy here, and without a key nothing generates.
- It runs the models in its catalogue, all through WaveSpeed: not every model WaveSpeed hosts, and no other provider.
- Carousels and each video mode run for one member of a workspace at a time. Single images run side by side for everyone.
- There is no public API to drive it from code: it is a web app.
- No model holds a face or a label perfectly every time, and results vary by model. That is what the review pass is for.