Rivya Chat Studio: Messages, Images, and Saved Chats

Chat Studio is Rivya's signed-in workspace for text conversations, writing, planning, and image analysis when the selected model supports images. You can continue saved conversations and use replies to prepare image or video tasks. Chat does not automatically execute a complete media project. For ways to organize longer work, see chat workflows.

Send your first message

1. Sign in and open the chat workspace from Studio. Select an available chat model; its controls and attachment options determine what you can use here. Check model availability if the model you expected is missing.

2. Write the task, relevant facts, and desired response format. For example: “Write three product-photo concepts for a blue ceramic mug. Keep the mug blue, use no invented product claims, and describe the setting and lighting for each.” This is a sample request, not a tested result.

3. Review the displayed credit hint and any model controls. If attaching an image, wait for the upload to finish and confirm the thumbnail. Send once; an image alone is not a message, so include a text request.

4. Wait for the reply to finish, read it, and follow up in the same conversation when you want the earlier exchange considered. The current message limit is 8,000 characters; split longer source material into purposeful parts rather than pasting an entire project at once.

A reply can help you draft or reason, but it does not verify the product facts you supplied. Review constraints and claims before using the answer. Prompt writing basics explains how to make a request specific.

Attach images and use image shortcuts

Image upload appears only for models that support attachments. The current chat uploader accepts JPEG, PNG, and WebP images up to 10 MB each; the number of images depends on the selected model. It does not accept PDFs, videos, or audio recordings as chat attachments. See references and uploads for other input routes.

Check the thumbnail and attachment count before sending. If you select more files than the remaining slots, only the files that fit are selected for upload. Uploading transfers the file before you send the chat message. Removing its thumbnail removes it from the pending message, not necessarily from storage; prepare a suitable copy using the safe upload guide.

When an image is available in the conversation or pending attachments, image shortcuts can analyze it or request an improved image or video prompt. These shortcuts send a chat request immediately, with the same credit implications as a message. Tool starter suggestions, when shown, instead fill the text box so you can edit before sending. See tools for those task presets.

Choose model controls and understand costs

Web search, reasoning effort, and thoughts controls appear only where the model supports them. Web search starts off. Enable it when relevant, then verify any cited page and its date yourself; a generated link or answer is not proof that a claim is current. Reasoning and thoughts settings differ between models and do not guarantee correctness or identical responses.

The displayed credit hint is a starting indication, not a fixed price for every turn. A chat turn can reserve credits using the selected model, conversation context, and expected response. Final settlement may return unused credits when the usage is confirmed; see credits and billing. Longer conversations can affect cost, and each follow-up or image shortcut is another request.

If a turn fails, use failed tasks and refunds to understand the difference between a registered failure and a completed answer you do not like. A disconnected browser alone does not establish either outcome.

Continue a saved conversation or start fresh

Successfully completed exchanges are saved to your account and can be reopened from History. The conversation retains its model and relevant saved controls. A visible streaming fragment or text still in the composer is not proof that a completed exchange was saved; see saved sessions.

New chat clears the current conversation view, unsent text, and pending attachments. It does not delete earlier saved sessions. Save important drafts separately before using it, refreshing, or changing pages.

Changing models can take you to a different model page or clear the current conversation. It does not migrate a saved conversation to another model with all of its context. Copy the facts and decisions you need into a new request, check attachment support again, and retain a copy of unsent work. Some current-page model switches carry draft input, but this is not a general autosave promise.

Take a reviewed prompt into image or video generation

Where available, Use for image and Use for video open the corresponding generation page with text from the reply. A nearby image reference may also be carried over; this does not transfer all conversation images or the full chat history. These links open the next form and do not submit a generation task.

If the reply has separate image, negative, and video prompt sections, the interface may display a prompt pack with copy controls. Review the relevant section before using it. Negative instructions are not guaranteed to populate a destination field, and long text carried by the shortcut is shortened. For a long brief, copy the complete approved version and compare it with the destination form.

Before generating, confirm the destination model, input mode, reference image, prompt, parameters, and credit estimate. The suggested destination is not a guarantee of model suitability. Follow Image Studio or Video Studio for submission. For audio, copy the appropriate script or music description into the audio workflow yourself.

Recover from an unclear or interrupted reply

Keep a copy of the request before retrying. Reopen the original conversation from History to check what was saved; do not infer completion solely from text that briefly appeared during streaming. If the answer exists, continue from it rather than paying to repeat the same request.

For upload or session problems, check browser and file requirements. If the issue persists, follow troubleshooting with the model, approximate time, and visible error. Share only the information needed to investigate, and review factual, numerical, and commercial statements before publishing a chat-generated draft.