Compare Rivya chat, image, video, and audio models by task, input mode, references, credits, quality settings, and availability.

Use this model selection guide when the question is not "which brand do I recognize?" but "which Rivya model fits this task?"

Rivya's AI Models catalog is the inventory layer for chat, image, video, audio, and tool-shaped work. Use it to compare task fit, supported modes, references, credit shape, and availability before spending credits.

The current catalog documents 131 model pages, while 86 models are currently available. Documented means the public page and capability record exist; it does not mean the model is available, directly runnable, or exposed through Public API v1.

If you are choosing models for a developer integration, use API Models and the API Model Reference alongside the catalog. Only models marked as API-available include runnable Public API examples; unavailable and Studio-only models stay useful for planning but should not be called from Public API v1.

Model Questions This Page Answers

Use this page when you need to decide:

which Rivya AI model fits a chat, image, video, audio, or tool task
whether the task needs text-to-image, image-to-video, audio cleanup, or another supported mode—and whether that mode has an available model today
whether reference images, source video, uploaded audio, or saved history should guide the run
how credit hints, quality settings, resolution, and duration affect the first model choice

Start With The Job

Before comparing models, write the job in plain language.

Examples:

create a clean product photo
animate an existing product image
generate a short product demo video
create a voice-over for a launch teaser
clean background noise from an uploaded recording, once an available cleanup model exists
reason through a math problem step by step
brainstorm and shape prompts in chat

The job determines which model fields matter. A model that is excellent for style exploration may be wrong for controlled product work.

Filter By Category

The model catalog groups work by broad category:

Category is only the first filter. It tells you the general medium, not the exact workflow.

Within the same category, models may differ by supported modes, reference inputs, quality settings, duration, speed, and credit behavior.

Check Modes Before Brands

Supported modes describe the model contract, but availability decides whether Rivya can run that contract now.

For example:

text-to-image is not the same as image-to-image
text-to-video is not the same as image-to-video
video-to-video is not the same as a simple prompt-to-video run
text-to-speech is not the same as audio cleanup
chat responses are not async generation tasks

If the mode does not match the job, brand familiarity will not help. If the mode matches but the model is unavailable, the page remains useful for planning but must not be turned into an execution recommendation.

Check References Early

Reference support can change the whole decision.

Ask:

does the model accept references?
what kind of references can it accept?
how many files can it use?
does the task need product identity, composition, style, or audio source control?

If references are central, read References and Uploads in Rivya and Image References in Rivya before running.

Use Credit Hints

Credit hints help you plan before spending.

They are not always a single final price because models can differ by:

fixed generation cost
duration-based cost
quality or resolution setting
token-based chat usage
provider-specific usage behavior

Use credit hints to compare likely cost shape, then check settings before submitting the task.

For wallet behavior, read Credits & Billing in Rivya.

Choose By Stage

Different stages need different models.

A rough pattern:

explore cheaply when the direction is unknown
use stronger reference control when structure matters
increase quality when the direction already works
use video only when motion is the deliverable
use audio after the script or scene is clear

The best model for final polish is not always the best model for the first run.

When A Model Is Unavailable

Model availability can change because of provider access, routing, configuration, or workflow readiness. Treat these states separately:

cataloged or documented: a public model record exists
available: the model can be considered for a current workflow
directly runnable: availability and direct-generation gates both permit execution on that surface
paused or temporarily unavailable: the page may remain public, but it creates no task and should not be presented as runnable

When a model is unavailable or not suitable, choose a fallback by task:

keep the same category if the medium is still right
match the supported mode first
match reference needs second
compare credit and quality settings third
only then compare style or provider preference

For more detail, read Model Availability in Rivya.

Model Selection Flow

Use this flow before spending credits:

Define the output type: chat, image, video, audio, or tool.
Decide whether references or uploads are required.
Compare supported modes on AI Models or model detail pages.
Check credit hints and settings.
Run a narrow first run.
Review the result before switching models.

Before step 5, stop if the detail page says Currently unavailable. Choose an available model with the same task shape, or state clearly that no direct replacement exists.

Do not change models before you know whether the prompt, reference, setting, or model caused the issue.

Rivya Model Selection Guide