Rivya Settings: Quality, Duration, and Aspect Ratio
Choose image and video settings in Rivya by model and destination. Understand resolution, duration, reference constraints, credit estimates and output checks.
Last reviewed on September 10, 2026
Choose settings for the result you need to deliver. Quality, resolution, duration and aspect ratio answer different questions; increasing one does not automatically improve the others. Rivya exposes controls by model and mode, so there is no single settings panel shared by every image, video and audio task. Start with model fields and parameters if a field is unfamiliar.
Distinguish the four settings
| Setting | What it controls | What it does not establish |
|---|---|---|
| Quality | A model-specific generation option, when offered | Correct product details, readable text or a better creative result |
| Resolution | Output pixel dimensions or a resolution tier | Accurate content or the same field of view |
| Duration | Requested output length, where supported | More complete storytelling or a universal audio length control |
| Aspect ratio | The relationship between frame width and height | Pixel count; a wider ratio is not a higher quality level |
Check available choices in the selected model’s form. Labels and combinations differ between models. For example, Rivya’s Seedance 2 Fast contract accepts 480p or 720p, while Seedance 2 accepts 720p, 1080p or 4K. Both accept requested durations from 4 to 15 whole seconds. These are specific model limits, not site-wide options; reference inputs can further restrict valid combinations. See the model directory and availability guide before choosing a model for a required format.
Set the frame before judging the composition
Choose the destination first: a wide website placement, a vertical mobile video or a square image may need different compositions. Treat these as layout examples, not fixed requirements for every platform. Check the actual dimensions and safe areas of your intended placement.
Cropping afterward can remove a label, a face or the main action. If one product asset needs both wide and vertical versions, review each composition separately. Leave room for text that will be added in editing; an aspect-ratio choice does not guarantee correctly rendered lettering.
References also affect framing. Some image-to-video requests do not send an independent aspect-ratio parameter. For example, Rivya removes that parameter for Kling 2.6 when references are attached. Do not assume a visible choice or a phrase in the prompt will override the reference-driven mode. Review image references or video references, and inspect the resulting frame.
Use duration and detail for a specific purpose
A short video can test one camera movement or one product action. A longer request needs enough meaningful action to fill the time; it does not make an overloaded script coherent. If the deliverable needs several distinct shots, plan them through video workflows and assemble them in an editor where needed.
For speech, check the text, pacing and the chosen model’s actual controls. Do not assume every audio tool has a duration selector or that entering seconds in the script creates one. Use audio workflows for the task-specific process.
Higher resolution can be useful when the destination needs more pixels. It cannot prove that a generated label, product shape or claim is correct. Likewise, upscaling is a separate operation where available, with its own input limits and cost; it is not a repair guarantee. Use Image Studio or Video Studio for the actual generation and result operations.
Check the complete request before spending credits
Write down the destination and one acceptance criterion, such as keeping the entire bottle and label visible. Select a compatible model and mode.
Attach required references, wait for uploads, then check the available quality, resolution, duration, ratio and audio options. Verify the files actually attached.
Read the current credit estimate. Change one relevant setting for a trial, then recheck the estimate before submitting. Do not assume a shorter request or lower setting reduces the total project cost after retries.
Watch, listen to or inspect the complete result against the criterion. Save a useful version and record its model, settings and prompt before making another change.
Cost is model-specific. Some requests use per-second pricing; some include reference-input duration; others reserve a fixed amount before final settlement. In the Seedance 2 reference-video flow, the estimate can include both requested output duration and uploaded reference-video duration. Changing only the output length therefore does not describe the entire bill. See credits and billing.
Each new generation is another request. A technically successful result that misses your creative expectations is not automatically refundable; see failed tasks and refunds. Settings do not provide a completion-time guarantee. For waiting or a connection interruption, check generation queues and waiting before resubmitting.
Recheck after changes and keep the accepted files
Changing models can reset parameters and remove unsupported or excess attachments. Recheck every field when changing the model, mode, reference material or destination. A prompt cannot enable an unsupported resolution, duration or ratio.
Before final use, verify actual pixel dimensions, framing, length, sound and the details that matter throughout the output. Compare product facts with the original material, not only with the previous generated version. Follow the commercial review checklist where applicable.
Download accepted outputs and keep your own files and settings. A saved result is useful evidence for the next attempt, but changing its settings and generating again does not guarantee the same scene with only that one difference.