Read Rivya model fields for categories, modes, direct generation, credits, references, uploads, quality, duration, aspect ratio, and outputs.

Use this guide when a Rivya model page has fields or controls that are unclear before you run a task.

Model fields tell you whether the model can run the job, what inputs it expects, which settings may affect cost, and how risky the first run might be.

Read Fields In This Order

When comparing models, use this order:

category
supported modes
direct generation status
reference and upload support
credit hint
quality, duration, aspect ratio, and output settings
strengths, examples, and related models

This keeps the decision grounded in what can actually run.

Category Is The First Filter

Category tells you the broad surface:

chat
image
video
audio

It is only the first filter. Two video models can support different workflows. Two audio models can have completely different forms. Two image models can differ sharply in reference capacity.

Use category to narrow the list, then inspect the fields below.

Supported Modes Tell You The Real Job Shape

Supported modes tell you what the model is currently meant to do.

Examples include:

text-to-image
image-to-image
text-to-video
image-to-video
video-to-video
text-to-speech
audio-to-audio
chat-completion or responses

Same category does not mean same mode. If the supported mode does not match the task, do not start there.

Direct Generation Answers Whether You Can Run It Here

Direct generation tells you whether the model can currently be run directly through the Rivya workflow.

If direct generation is not enabled, the model may still exist in the catalog, but it is not the right place to expect a live run.

Use this field before spending time on a prompt.

Credit Hint Is A Planning Tool

Credit hint helps you compare likely cost before submitting.

Read it as a planning signal, not a universal final price.

Rivya models can broadly fall into these billing shapes:

fixed-cost generation
duration-based generation
quality or resolution-dependent generation
token-based chat usage
provider-specific metered usage

Before a heavier task, also check Credits & Billing in Rivya.

References And Uploads Change Model Choice

Reference support can matter more than model popularity.

Check:

whether references are supported
what file types are accepted
how many files can be used
whether the model expects prompt-first or upload-first input
whether the reference is for product identity, composition, style, motion, or audio source

A single reference can help with a visual anchor. Larger reference limits can support product packs, style systems, and consistency work.

For more detail, read References and Uploads in Rivya.

Aspect Ratio And Image Size

Aspect ratio or image size controls output shape.

It matters when:

the image must fit a product page
the output needs a social crop
a landing-page hero needs wide framing
a short video needs vertical placement
composition would break if cropped afterward

Choose shape before generation when placement is known.

Quality And Resolution

Quality and resolution often affect both finish and cost.

Use lower or standard settings when:

testing a prompt
comparing models
validating references
exploring style

Use higher settings when:

the direction already works
the asset is close to final use
the placement justifies more detail

Do not spend high-quality credits proving a vague idea.

Duration

Duration matters for video and some audio workflows.

Shorter duration is better for:

motion tests
social clips
quick voice checks
proof-of-direction runs

Longer duration is better when the script, story, or demonstration actually needs time. If a video feels overloaded, split it into multiple assets instead of stretching duration.

Audio And Sound Controls

Some video and audio models include sound-related controls.

Those controls can change the type of output, not just the style.

For example, a silent motion test, a video with native audio, a voice-over, and a sound-effect cue are different tasks. Choose controls based on the deliverable.

Form Shape Explains Why Pages Look Different

Rivya model pages can expose different form shapes:

prompt-first generation
text-to-speech forms
dialogue-style forms
upload-first cleanup forms
chat forms
reference-led generation forms

That difference is intentional. The form reflects what the model needs to run.

A Practical Pre-Run Checklist

Before submitting a model task, check:

Does the supported mode match the job?
Is direct generation available?
Does the task need references or uploads?
Is the credit hint acceptable for the stage?
Are quality, duration, and aspect ratio appropriate?
Is the prompt narrow enough to review?
Do you know what would make the output usable?

If any answer is unclear, start with a smaller test.

Rivya Model Fields and Parameters Guide