↓
Rivya AI Docs

Rivya Chat Workflows: From a Brief to a Usable Draft

Plan writing, research, calculations, and image or video briefs in Rivya. Choose a chat entry, check the answer, and preserve the decisions needed next.

Last reviewed on November 11, 2026

Use chat when the next useful result is a decision, explanation, draft, or prompt. Start directly in a generation workspace when you already have an approved brief and need a media file. Chat can help prepare that brief, but it adds its own requests and costs; it is not a required step or a guaranteed saving. For sending messages and using the current chat controls, use the Chat Studio guide.

Text-only Gemini 4 Carbon requests

Gemini 4 Carbon accepts text without attachments, web search, tools or exposed reasoning settings in this Rivya path. It delivers one complete answer after terminal validation. The reserve covers the model’s full published context for both input and output; it does not use a default answer length as a billing cap.

The current promotional quote expires at 00:00 UTC on November 30, 2026. New requests stop then until a new quote is confirmed. Existing requests keep their creation-time quote and settle from confirmed consumption. Do not automatically continue or resend an uncertain request.

Choose where to start

What you needStarting pointWhat to check
Explore an idea, rewrite copy, or compare approachesA regular chat from StudioState the audience, facts, constraints, and desired answer
Calculate a quantity or solve a defined reasoning problemAI Calculator or AI Solver through ToolsProvide the complete problem, units, and assumptions
Prepare a visual brief from a reference imageText-only chat with your written observationsInspect the image yourself and supply details you can verify
Continue a previous decision or draftThe original conversation in HistoryRead the saved exchange before asking for the next change

Public chat, model, and tool pages help you compare options. Executing a chat request requires an account; a public page is not anonymous free execution. See public and signed-in workflows. If a model cannot be selected, check model availability.

Turn a vague task into a usable brief

  1. Describe one outcome and who will use it. For example: “Prepare a product-demo outline for a blue ceramic mug, for first-time shop visitors. Use only these confirmed facts: ceramic material and blue color.” This is a writing example, not a tested generation result.

  2. Ask the model to separate confirmed facts, open questions, and proposed creative choices. Do not let an invented capacity, dishwasher claim, or shipping promise become a product fact. Supply missing information yourself or leave it explicitly unresolved.

  3. Choose one direction and request a concrete draft: scene order, approved wording, and a separate visual description. Ask for a targeted revision such as “Remove the durability claim; keep the scene order,” rather than repeatedly asking for something “better.”

  4. Read the final draft against the original brief. Keep the accepted text, facts that must stay unchanged, remaining questions, and intended next step in your own project notes before moving on.

You can follow up in the same conversation to use its earlier context. That context may also contain rejected ideas, so explicitly identify the version you accepted. Prompt writing basics covers the structure of a clear request.

Check research and calculation answers

For current information, check sources outside Rivya and include relevant excerpts and their dates in your prompt. Chat models currently available in Rivya do not offer web search. Ask the answer to distinguish the source text you supplied from assumptions; that instruction does not make the answer verified.

AI Calculator and AI Solver are chat tools with task instructions and starter suggestions. They do not turn a language-model answer into a certified calculation. For a small problem, independently check the result. For example, 24 bottles × 750 mL = 18,000 mL = 18 L; for 3x + 5 = 20, x = 5, and substituting 5 gives 20. These are hand-checked examples, not claims about a live tool run. A correct total can still rely on a wrong input or unit.

The tool context belongs to the conversation. Start a new, appropriate conversation if the task changes substantially instead of assuming a saved tool session has become an unrelated general-purpose project.

Prepare image, video, or audio work

For visual planning, inspect the reference yourself and describe the relevant visible details in text. Use reference uploads in an available image or video operation that supports them. An image does not establish hidden product specifications, ownership, or permission to use a person or brand. Follow references and uploads and the safe upload guide. The current chat composer does not accept images, PDFs, audio recordings, or video files.

For an image, keep the subject, composition, and facts to preserve explicit. For a video, also describe action, camera movement, and the intended sequence. If a reply offers an image or video shortcut, it opens the destination without generating. It can carry the selected prompt and an image reference from an older conversation when one is available. Review the restored inputs and keep a complete approved brief in your own notes; do not assume full chat history or every attachment will transfer.

Choose an available destination operation, then check its mode, accepted references, supported prompt and settings, and credit estimate before submitting through the image workflow or video workflow. For narration or music, manually copy the appropriate script or description where an available operation accepts it, using the audio workflow. A plan in chat does not assemble or synchronize those media files. The guide to working across Studios provides broader context.

Keep work recoverable and costs visible

Successfully completed exchanges can be reopened in the account's saved conversations; generated media is tracked separately as tasks. Neither is a complete project archive. Keep accepted briefs and required source files outside the current form as well. See saved sessions for what is persisted; unsent text and a partially streamed answer are not proof of a saved exchange.

Every follow-up is another chat request. Longer conversation context can affect reserved credits and final cost, while image, video, and audio generation has separate charges. Check credits and billing before planning several revisions. A longer planning conversation does not necessarily cost less than generating directly.

Changing models or opening a new chat does not migrate all previous context. Preserve a concise, reviewed summary yourself, and verify the next conversation has the facts it needs. Save unsent drafts before refreshing or navigating away.

Decide whether to revise, generate, or stop

Revise in chat if an important fact, instruction, or output format is still wrong. Move to generation when the brief is reviewed and the next deliverable really is a media file. Stop when the requested text or decision is already usable; there is no need to add another model call just to follow a workflow.

If a reply is interrupted, check the original saved conversation before resending. A browser disconnect alone does not prove failure. If the saved answer exists, continue from it; otherwise use troubleshooting with the model, approximate time, and visible error. Before publishing any draft, verify its facts, numbers, and commercial claims.