An AI video generator that hands you an editable clip, not a download
Twenty-one models from ten labs behind one prepaid credit balance. Type a prompt, pick a model, and the result appears on your timeline as an ordinary clip you can immediately trim, grade and cut against.
The usual AI video workflow is a chain of tabs: subscribe to one lab, generate, download an mp4, open a separate editor, import, discover the shot doesn't cut against the previous one, go back. Bernini collapses that into a single tab. The Generate panel lives inside the editor, and a finished generation drops onto the timeline in the same state as footage you imported yourself — no locked render, no re-import.
The models
Video, with the starting credit cost per generation:
| Model | Lab | From |
|---|---|---|
| Veo 3.1 | 48 credits | |
| Veo 3.1 Fast | 24 credits | |
| Sora 2 | OpenAI | 24 credits |
| Kling 3.0 Pro | Kling | 34 credits |
| Kling 3.0 Standard | Kling | 26 credits |
| Kling 2.5 Turbo Pro | Kling | 20 credits |
| Seedance 2.0 | ByteDance | 92 credits |
| Wan 2.5 | Alibaba | 15 credits |
| Happy Horse 1.1 | Alibaba | 42 credits |
| Hailuo 02 | MiniMax | 17 credits |
| Grok Imagine | xAI | 15 credits |
| Ray 3.2 | Luma | 30 credits |
| PixVerse V6 | PixVerse | 8 credits |
Image, for stills, thumbnails, backgrounds and image-to-video source frames:
| Model | Lab | From |
|---|---|---|
| FLUX.2 Pro | Black Forest Labs | 3 credits |
| FLUX 1.1 Pro | Black Forest Labs | 3 credits |
| FLUX.1 Dev | Black Forest Labs | 2 credits |
| FLUX.1 Schnell | Black Forest Labs | 1 credit |
| Nano Banana Pro | 9 credits | |
| Nano Banana 2 | 5 credits | |
| Seedream 5.0 Pro | ByteDance | 5 credits |
| GPT Image 2 | OpenAI | 13 credits |
Costs scale with duration, resolution and whether you ask for generated audio. The Generate button prices the exact combination before you run it, so it never quotes a number the server won't honour.
Five ways to generate
- Text to video — a prompt in, a clip out, on any of the 13 video models.
- Image to video — attach a still and Bernini routes to that model's image-to-video endpoint automatically. Supported on 11 of the 13.
- Character to video — save a Character once and reuse the same subject across shots, on Seedance 2.0, Grok Imagine and Veo 3.1 Fast.
- Text to image — stills, thumbnails, backgrounds, overlay plates.
- Image editing — send an existing image plus an instruction to FLUX.2 Pro, Nano Banana Pro or 2, Seedream 5.0 Pro or GPT Image 2.
Why one balance instead of many subscriptions
Model quality moves month to month, and the model that nails a particular shot is rarely the one you already pay for. Subscribing to each lab separately to find out is expensive and slow to cancel. Credits are prepaid and flat-rate across every model, so comparing Veo against Kling against Seedance on the same prompt costs credits rather than three recurring bills. Pro includes 300 credits a month — roughly ten video generations — and you can buy more without changing plan.
If you'd rather not use credits at all, you can bring your own API key and run generations against your own provider account.
And then you still have to edit it
Generated clips rarely cut together on their own — they need trimming, speed matching, colour work and a transition between shots. Because generations land as normal timeline clips, the full editor applies: 67 GPU-accelerated effects, 33 transitions, keyframe animation, speed ramping, AI captions and hardware-accelerated export up to 1440p. See the browser video editor for what that covers.