> ## Documentation Index > Fetch the complete documentation index at: https://kive.ai/docs/llms.txt > Use this file to discover all available pages before exploring further. # Chat mode > Iterate on any generation through conversation — describe edits, chain tasks, and generate video Chat mode lets you iterate on any generated image through conversation. Describe edits in plain language, chain multiple steps together, or run tasks in the background while you work on something else. It's the fastest way to go from a first generation to a polished final result. ## Where to find it Open any generated image in the **Create** section. Chat mode appears as the **Edit** panel on the right side of the screen. You'll see the original prompt, quick actions like **Turn into video** and **Zoom out**, and a text field where you can describe what you want to change. Chat mode panel showing the Edit tab, original prompt, quick action buttons, and text input field ## Describe what you want Write in natural language. Chat mode understands requests like "make it a close up", "remove the object in the corner", or "add dramatic lighting". The agent thinks through your request, then generates a new version. Each edit costs credits and builds on the conversation history, so you can refine results step by step. Chat mode conversation showing a request to change the jacket to a different product, with the generated result You can also click the paperclip icon to attach reference images — useful for style references, start and end frames, or any visual context you want the agent to consider. ## Reference custom models Type `@` in the text field to mention any custom model from your workspace. The model gets applied to your request, so you can swap products, change outfits, or add accessories in a single message. Dropdown list of custom models appearing when typing @ in the chat input For example, you might type "Change the jacket to @Fiction -- Hoodie" and Chat mode will replace the jacket with that specific product while keeping everything else intact. This works for any model type — products, characters, or styles. ## Generate video Turn any image into video by describing the motion you want. Ask for camera movements like "slowly push in on the product", subject animation like "make this person walk around", or combine both. Chat mode will show a confirmation step with the credit cost before generating. Chat mode video confirmation dialog with Cancel and Create video buttons Common motion prompts include orbit shots ("do a 360 around it"), camera pushes ("push in with a cinematic vibe"), and subject movement ("make them take a spin while the camera pulls back"). ## Chain tasks and run in the background You can ask the agent to run a sequence of steps in a single conversation. Generate different angles from your shot, then create videos between those angles — all without leaving the chat. When you start a task that takes time, you don't need to wait. Open a different image or start a new conversation, and your previous tasks keep running in the background. The notification badge shows how many sessions are active. Background agents panel showing three chat sessions in progress alongside completed sessions ## High-fidelity video with start and end frames When you need precise control — especially for products with text, logos, or fine details — you can define both the start and end frames to avoid morphing artifacts. **Generate the end frame first.** Ask Chat mode to create what the final frame should look like, then request a video transitioning between the two. For example: "First generate an image where the camera has moved closer and flowers have appeared on the sides. Then generate a video transitioning from the original to that new image." Chat mode conversation showing a chained workflow — first editing the image, then requesting a video with start and end frames Here's what the resulting video looks like — the camera pushes in smoothly while foliage grows into the frame, transitioning from the original to the generated end frame: