The model list shows each model's description, its starting price and what it supports: **Image Ref** for reference images, **Frames** for start and end frames, **Video Ref** for reference videos. Use the search box to filter by name. The default model is a good starting point for most prompts. Models differ in style, detail, text rendering and speed, so trying the same prompt on two or three models is the fastest way to find the one you like. Models marked **New** were added recently.
- **Ratio** sets the shape of the image or video. The ratios available depend on the model. **Auto** lets the model decide. - **Resolution** sets the output size and is the main factor in the price. - **Quality** appears for models that offer several quality levels; higher levels cost more credits. - **Duration** and **Sound** apply to videos. Videos are charged per second. You can also write these in the prompt: for example 16:9, 2K, 1080p or 5s. The settings follow the prompt when the model supports that value.
Add reference images with the attach button in the prompt box, by dragging files onto it or by pasting an image. Supported formats are PNG, JPEG and WebP up to 20 MB each; the number of references depends on the model. Type @ in the prompt to point at a specific reference by name, for example to say which image is the subject and which is the style. On a result's page, **Remix** sends it to Create as a reference, and **Reuse Prompt** brings back its prompt and references.
AI Enhance rewrites your prompt with a language model to add detail about the subject, scene, lighting and style while keeping your intent. - With **Auto-enhance on send** turned on, prompts are enhanced automatically when you press Generate. - **Enhance now** rewrites the text in the prompt box so you can read and edit it first. **Undo** restores your original text. Results show the prompt you wrote; an **AI Enhance** label on the result opens the enhanced version that was actually used.
Open the image and choose **Editing**. Draw a box or click on the areas you want to change; each mark is added to the prompt as a numbered thumbnail. Describe what should happen to each marked area, then send. You can place up to 10 marks on an image, combine marks with reference images, and the result appears in Create like any other generation. The original image is kept.
Open an image and choose **Animate** to send it to Create as the reference for a video model. With a video model that supports frames, turn on **Frames** in the toolbar to set a **Start** and an **End** image; the video moves from the first to the second. Models with **Video Ref** accept reference videos (MP4, WebM or MOV up to 30 MB each) instead. Open a video and choose **Editing** to trim a clip on the timeline and mark areas on specific frames; the selection is used as the reference for the next video.
Prompts and reference images are checked before they reach a model. ShotPoet does not generate: - sexual or suggestive content, and any sexual content involving minors; - face swaps, deepfakes or other edits of a real person's face, body or identity, and realistic depictions of named real people in fabricated or damaging situations; - violent, hateful or harassing content; - content that helps with illegal activity or fraud. Changing the background, lighting or colors of a photo of a real person, or adding objects to it, is fine. The full rules are in the [Acceptable Use…