↓
Rivya AI Docs

Rivya Glossary: Credits, Models and Task States

Look up Rivya terms for Studio, models, tools, credit reservations, task states and outputs. Understand what each term means before starting a task.

Last reviewed on November 5, 2026

Use this page to look up a term you see in Rivya's interface, pricing or guides. Product names such as Studio and model names keep their original spelling; explanations describe how they work in Rivya.

Pages and saved work

  • Public pages: pages you can browse before signing in, including model introductions, tools, pricing, blog posts and documentation. Browsing a page does not submit a paid request.

  • Public start page: a public page where you can explore a workflow and, when a form is provided, prepare a request. Sending a chat message or generation requires sign-in; a supported public form can submit after sign-in without a move into Studio. See public and signed-in workflows.

  • Studio: the signed-in work area for chat, images, video and audio. It is not a promise that every unsent draft or setting is saved. See Studio.

  • Dashboard: an account overview with your balance and a limited selection of recent work, rather than your complete history.

  • History: saved conversations and generation records belonging to your account. Chat and tool conversations have their own history; image, video, audio and generated-text tasks appear in generation history. The list shows saved task states; opening it does not poll each task. A task record does not guarantee its output file will remain downloadable. See History.

  • Notifications Center: account and task events, such as payment updates or generation outcomes. An event is not a substitute for the task's current status or your billing records.

  • Settings: account-management pages. /settings/billing shows subscription information; /settings/credits shows credit balances and transactions.

Credits and charges

  • Credits / wallet balance: the units you can spend through your Rivya account. Billable chat, generation, tool and API requests use the same account balance. Credits are not a separate cash balance.

  • Signup credits: a one-time starter grant for a new account after email verification, when the signup gift is offered. It has an expiration date and is not a recurring monthly allowance on the Free plan.

  • Subscription credits: credits issued monthly with a paid subscription. Annual billing does not mean the entire year's credits arrive at once.

  • Credit pack: a one-time credit purchase with its own validity period. Buying a pack does not start a recurring subscription.

  • Spending order: unexpired credits with the earliest expiration date are used first. For equal expiration dates, earlier grants are used first. Credits without an expiration date come after dated credits. “FIFO” alone is therefore not an accurate description of which credits Rivya spends first. See credits and billing.

  • Credit estimate / credits hint: an indication of cost. Check the estimate after choosing the model and settings; a general model-page hint is not a final charge.

  • Reserved credits: credits deducted before a billable request is sent for processing. Settlement compares valid reported usage, rounded up to whole credits, with that reservation. When the amount is at or below the reservation, it becomes the final charge and any difference is returned. Eligible fixed-price Runway and Suno Sounds generation requests can also settle missing usage at the reservation when it matches their recorded quote. Other missing, invalid or above-reservation usage leaves the reserved amount charged while settlement is checked; a completed result alone does not confirm a final charge.

  • Credit return: credits added back to the Rivya balance, for example after a qualifying failure or unused reservation. It is not a refund to your payment card, and a partial return can occur on a successful task. See failed tasks and credit returns.

Models, tools and inputs

  • Model: the AI system selected for a supported task. Capabilities and input requirements vary by model, even within one category.

  • Model page: a public introduction to a catalog entry. It may describe a model that is not currently runnable in Rivya.

  • Availability / direct generation: availability describes a model's current access state; direct generation indicates whether Rivya enables new requests through its supported workflow. A catalog listing alone does not enable a request. For Public API use, also check the model's API status and supported inputs. See model availability.

  • Tool: a guided chat entry, such as AI Calculator for calculations or AI Solver for worked problems and explanations. It supplies preset instructions to the selected chat model and uses account credits; it is not a separate subscription. See tools.

  • Prompt: the instruction or description you give a model. For a text-to-speech workflow, the main text is the script to speak; extra instructions in that script may be spoken too. Use the selected form's separate controls for supported voice settings.

  • Reference / upload: a file supplied to a workflow that accepts it. A reference can guide a result; other supported inputs may be transformed or analyzed. The selected model and workflow determine file type, count, size and any duration or dimension limits. Sign in to use supported upload controls, and check that the model accepts the file before submitting. An uploaded file is not usable by every model. See references and uploads.

  • Parameter: a supported setting such as duration, aspect ratio or resolution. A setting on one model may not exist on another. See model fields and parameters.

Tasks and results

  • Task / generation request: a submitted request to create an image, video, audio or supported text result, such as lyrics. Chat messages belong to conversations and use a separate turn flow. Submitting another generation creates another request that may incur another charge, even with the same prompt.

  • Task lifecycle: the stages from checking inputs and handling credits to processing and recording an outcome. See the task lifecycle.

  • WAITING: the task is waiting; this is not a final outcome.

  • GENERATING: Rivya is waiting for a result or confirmation of the request. This can include a model-service queue or an unconfirmed submission. A long wait alone does not establish failure.

  • SUCCESS: the task has a recorded successful outcome. This does not guarantee the content meets your brief, the credit charge is final, or the file remains accessible forever.

  • FAILED: a recorded failure. Read the error and credit transactions before trying again.

  • Callback / polling: ways Rivya receives or checks a provider's task result. You do not need to inspect these mechanisms to ask for help.

  • Output: the generated response, text or media file. Available media downloads open a file URL that you can save through your browser; saved lyrics have a text download. Saving a task record and keeping a local copy are different actions. See downloading outputs.

Providers, rights and private information

  • Provider: a third-party service used for an enabled function, such as generation, payments, sign-in, storage or email. Which provider processes information depends on the workflow you use. Upstream provider refers to a service involved in processing the request behind Rivya's interface; it is not an additional workspace you must open.

  • Commercial use: using a result in a business context, such as a client project, advertisement or product asset. The label alone does not confirm that a particular result is cleared for that use.

  • Output ownership: the allocation of output rights described in Rivya's Terms of Service. Read the terms and relevant provider conditions; this glossary does not grant additional rights or guarantee uniqueness or freedom from third-party claims.

  • Sensitive data: personal, confidential or otherwise sensitive information in prompts and files. Review what you submit and which service processes it. Rivya's Privacy Policy describes data handling; see data and provider processing for workflow guidance.

If a term is clear but the action still fails, use troubleshooting to find the relevant record and the information to send support.