# Generate with AI

> Images, video, music, sound effects and voice from hosted models, priced before you start and added to your project.

Formats Studio docs · Creative tools · https://formats.app/docs/generate · Updated 2026-09-28

## Open Generate

Open **Library** › **Generate**, or choose **Generate with AI** from the timeline dock. The panel has four modes: **Image**, **Video**, **Audio** for music and sound effects, and **Voice**.

## Make something

1. Pick a mode, then a model from the picker. Each model shows who made it, what it is good at and its price.
2. Add references where the model takes them: image references for image models, a **Start frame** and an **End frame** for video models that animate between them.
3. Describe the shot: subject, action, camera and mood. In an empty prompt, type `/` to reuse a recent prompt.
4. Adjust the settings the model offers, such as aspect ratio, resolution, length, variations or sound.
5. Check the estimate, shown as **≈ N credits** beside your balance, then press **Generate** or `⌘↵`.

The job appears under **Recent** with its progress. When it finishes, each result offers **Add** to put it on the timeline and **Save** to keep it in Assets.

> **Note:** Generation runs on hosted models and keeps running if you switch projects. The price you see before you press Generate is the price you pay.

## The models

| Mode | Models | Tier |
| --- | --- | --- |
| Video | Gemini Omni Flash (Google), Wan 3.0 (Alibaba) | Standard |
| Video | Veo 3.1 Fast (Google), Kling 3.0 Pro (Kuaishou), Seedance 2.0 (ByteDance) | Premium |
| Image | Nano Banana 2 (Google), Seedream 5.0 Pro (ByteDance), FLUX.2 Turbo (Black Forest Labs), Remove background, Upscale | Standard |
| Image | Nano Banana Pro (Google), GPT Image 2.5 (OpenAI) | Premium |
| Audio | Lyria 3.5 (Google), Sound effects (ElevenLabs) | Standard |
| Audio | Eleven Music (ElevenLabs) | Premium |
| Voice | Sonic 3 (Cartesia), 10 voices in 16 languages | Standard |

Every model runs on the credits that come with Creator. Premium models cost more credits per run, and each shows its exact price before you generate. See [Plans and credits](https://formats.app/docs/plans) and the current prices on [Pricing](https://formats.app/pricing).

## Other ways in

- From an image in Assets: **Generate with AI** to animate it into a video, edit it with a prompt, remove its background or upscale it.
- From a clip on the timeline: animate it into a video, edit it with a prompt, or animate the frame at the playhead.
- From the chat: ask your agent. It lists the models, proposes one with its exact price and waits for **Approve and generate**; results land in Assets for it to place.

## Related

- [Plans and credits](https://formats.app/docs/plans): What is free, what credits pay for, what Creator adds, and how credits are granted, spent and topped up.
- [Work with your agent](https://formats.app/docs/agent): Choose Claude Code, Codex or GitHub Copilot, ask for edits in plain words, and review, apply or undo what it does.
- [Local AI and transcription](https://formats.app/docs/local-ai): What runs on your Mac and what runs hosted: Local AI for your footage, and transcription.