# Create > Make images, videos and speech in Endless with Image, Video, Canvas and Speech, and learn when to use each one instead of a chat conversation. Source: https://endless.okinawa/docs/create The **Create** surfaces are for when you already know what you want, which model should make it and at what size. If you are still deciding, start in [chat](/docs/chat): you describe the result and Endless picks the model. A prompt, optional reference pictures and every setting the chosen model offers. The same composer for clips, with length as a setting and higher costs. A shared board you generate onto, when placement matters. Text read out loud in a voice you pick, on a canvas or in chat. **Image** and **Video** each have two tabs: **Manual** (you set the model and every control) and **Agent** (your prompt goes to a new chat that makes the result for you). Everything you make on any surface is saved to the same Library. See [Where generated output is saved](/docs/create/where-output-lands). ## Before you start * **A workspace can switch off** image generation, video generation and Canvas, each separately. A surface that is off is missing from the sidebar, and a run request is refused with *"Image generation isn't enabled for this workspace"*. Ask whoever runs the workspace. * **Generating needs permission to create content.** A role without it can open a canvas only as **"Read-only"**. See [Roles and permissions](/docs/governance/roles-and-permissions). * **Every run spends credits**, and the cost varies widely by model: one video can cost many times an image. The composer shows the estimate before you run. See [Credits and spending](/docs/concepts/credits-and-spending) and [Models](/docs/concepts/models).