Chat & Creative Tools

Sogni Studio ships with an in-app Chat panel — a creative assistant that can read your images, watch your videos, write prompts and lyrics, and call into the same generation pipeline you use directly. Conversations and the media they generate land in the same gallery as the rest of your work.

Chat is powered by Sogni Intelligence — Sogni's own LLM API running on the Sogni Supernet. There's no third-party model provider, no separate signup, and no data leaving the Supernet. You spend the same Spark balance.

#What it can do

Studio's chat assistant has access to a suite of tools — discrete actions the model can call to manipulate your work:

#Vision

• analyze_image — describes, tags, or critiques any image in your gallery or from your Mac.
• analyze_video — full parity for video: describes content, motion, scene structure, suggested edits.

#Generation

• generate_image — drafts a prompt and renders it. Useful for "show me something like X."
• animate_scenes — turns a single image into a short animated scene, with per-model size-preset resolution.
• generate_transition — generates a video transition between two clips. Use it to bridge cuts in Clip Mixer.
• extend_video — extracts the last frame of a clip and continues from there, automatically stitching the new segment.

#Writing

#How it works

The chat panel runs alongside the main Studio window. Type a request — "describe the mood of this image," "make a transition from clip A to clip B," "write a chorus for this verse" — and the assistant decides which tool (if any) to call. Tool calls execute in the background; their results land:

• Text appears in the chat thread.
• Images appear in the chat carousel and in the main gallery.
• Videos save to the gallery and to the Clip Mixer pool if the call was bridging clips.
• Audio lands as a normal gallery item.

Use the chat carousel to flip between alternate generations the assistant produced for a single request.

#Underlying LLM

The chat is powered by Qwen 3.6 running on Sogni's intelligence service. The tool-calling layer follows the OpenAI tool-use protocol, so the same suite of tools is available to your own apps via Sogni Intelligence. What Studio exposes interactively, you can call from code.

#Where to find it

Chat lives in Studio's right-hand sidebar as one of the sidebar's modes — switch the sidebar to Agent / Chat mode to bring it up. The chat panel works alongside any other Studio mode, so you can generate on the main canvas while the assistant analyzes media in the side panel.

#Cost

Chat consumes Spark for:

• Each LLM message (small).
• Any generation tool call (same cost as running that generation directly).
• Vision tool calls on images and videos (small, proportional to size).

Token use and tool-call costs are surfaced in the chat header so you can see what each turn is spending.

#Tips

• Show, don't tell. Drop images into chat as context — the assistant uses analyze_image automatically and writes much stronger prompts off a concrete reference.
• Ask for variants, not just renders. "Generate three alternate compositions of this scene" produces a chat carousel; the standalone generate panel is for direct one-shot work.
• Start a new conversation when you change direction. A fresh thread keeps the assistant's tool context clean; long single threads start mixing earlier intents into later answers.
• Copy out the screenwriter output. When the assistant writes a great cinematic prompt, copy it into Video Create mode rather than re-rolling — chat tool calls don't always reproduce exact prior outputs.

#See also