Rivya Model Fields and Parameters Guide
Check Rivya model availability, required files, settings and credit costs. Learn how website form fields differ from API request fields.
Last reviewed on September 12, 2026
A model’s category tells you where to look; its supported inputs, current availability and form tell you whether it can do your particular job. Start with the result you need, then read the fields below. You do not need to learn internal parameter names to use the website.
Check availability and the task type first
The model catalog groups models into chat, image, video and audio. A shared category does not mean shared capabilities: generating a video from text, animating a photo and driving a character with a motion reference require different inputs.
Read the current availability notice before preparing a prompt. A catalog page can remain visible while direct generation is unavailable. A listed capability or an example is not proof that the model can currently run in Rivya. See model availability.
Supported modes describe the input-to-output task, such as text to image, image to video, text to speech or dialogue to audio. The form may choose a mode from the files you attach and the options you select; do not assume there is always a separate mode selector. For a comparison of suitable starting points, use choosing models.
Read the input requirements before writing the prompt
| What to check | Why it matters |
|---|---|
| Prompt or spoken text | A visual description and text that will be read aloud serve different purposes. Follow the selected form and its length limit. |
| Required files | Some tasks start from text; others need an image, video, audio file or a specific combination. A prompt cannot replace a required upload. |
| File limits | Check format, size, count and any dimension or duration limits for this model and mode. |
| Reference role | A first frame, last frame, subject image, motion video and voice recording are not interchangeable. Use the appropriate field. |
| Additional controls | A dialogue form can assign voices to lines; another audio form may take a single text. Options differ by model. |
Wait for uploads to finish and confirm that the intended files are attached. A file selected on your computer is not yet evidence that it will be included in the request. See references and uploads and audio uploads.
A higher reference limit gives you room for more inputs; it does not guarantee product identity or consistency. Conflicting references can make the intended result less clear. Give each reference a purpose and compare the output against the original product or source material, not just a previous generated image.
Check which settings apply to this request
Aspect ratio describes frame proportions; resolution describes pixel dimensions or a resolution tier. Quality is a model-specific option, and duration is the requested output length only where supported. They are separate controls. Read quality, duration and aspect ratio for concrete model examples and their limits.
Some fields are hidden for particular modes, and a reference-driven request can omit an independent aspect-ratio parameter. Writing an unsupported value in the prompt does not enable it. Use the options accepted by the selected form and inspect the actual output.
Sound controls also have different jobs. Generating sound with a video, reading a script aloud and using an uploaded recording as an input are different workflows. A sound toggle is not a voice-upload field, and attaching audio does not promise that every video model will preserve or mix it. Consult video workflows or audio workflows.
Read the current estimate, then inspect the settled cost
A model’s credit hint is a starting point, not a final quote for every possible request. Cost can depend on the model, duration, resolution, input files or chat usage. Some tasks reserve credits first and settle the actual charge afterward. Changing a setting or reference can change the estimate; check it again immediately before submitting.
Include retries in your budget. A shorter test or lower setting does not necessarily make the finished project cheaper, and a technically successful result that you dislike is not automatically refundable. See credits and billing and failed tasks and refunds.
Recheck the form when switching models
Switching models can restore default parameters and remove unsupported or excess attachments. Review the prompt, files and every relevant setting again. Do not assume the new model inherits the previous request unchanged.
Before submitting:
Confirm that the current model is available and supports the task you need.
Check required inputs, completed uploads and the selected settings together.
Read the current credit estimate and define one acceptance condition, such as keeping the whole product label visible.
After generation, inspect the complete image, video or audio against that condition and save the usable result before trying another request.
Keep your own copies using output downloads. Higher settings do not guarantee factual accuracy, readable lettering or an otherwise identical scene when you regenerate.
Website fields and API fields are different instructions
This guide explains how to read the website’s model forms. For an integration, use the selected model’s page in the API model documentation, which provides its API model ID, supported inputs, request fields and readiness information. Do not copy a visible website label into an API request and assume it is a valid parameter.
Website availability and API input readiness need separate checks. A public model page or website upload feature alone does not prove that the same input path is available through the API.
GPT Image 2.5 Flare / GPT Image 2.5 Sunburst
Generate images from text or edit with up to 16 reference images. Choose 1K, 2K, or 4K and an automatic, transparent, or opaque background.
GPT Image 2.5 Flare · GPT Image 2.5 Sunburst
Gemini Omni Video / Gemini Omni 1.1 Flash
Create video from text, reference images, or a source video. Choose 16:9 or 9:16 and 4, 6, 8, or 10 seconds; with video input, the model determines output length.
Use up to seven images at 20 MB each, or one video up to 95 MB and 30 seconds with up to five images. Select a video segment no longer than 10 seconds. Audio and character asset IDs are not supported.
Gemini Omni 1.1 Flash: Frames mode accepts one start image or a start/end pair, without other references.