Rivya Image Studio: Generate, Review, and Save Images
Use Rivya Image Studio to create images, work from references, remove backgrounds, upscale, or separate layers. Check model settings, credits, and saved results.
Last reviewed on September 9, 2026
Image Studio is the signed-in workspace for still-image tasks. Use it for product concepts, listing visuals, ads, social graphics, or images for later video work. It records submitted tasks and their results; it does not automatically save a complete editable project or an unsent draft. See saved sessions for that distinction.
Choose the operation before the model
| Your task | What to choose | What to check |
|---|---|---|
| Create a new image from a description | A model supporting text-to-image | Prompt, composition, supported output settings |
| Create or edit with visual references | A model supporting image inputs | Required images, upload limits, what each reference should influence |
| Remove a background or upscale an existing image | A dedicated operation such as Recraft Remove Background or Topaz Image Upscaler | One eligible source image and that operation's controls |
| Separate elements from a flattened image | Seedream 5.0 Pro Layer Decomposition | One eligible source, optional guidance, and returned layers |
Browse the image overview and model directory while comparing options. Open an available model in Studio to upload and submit under your account. Public model pages do not provide anonymous execution. Check model availability if an option cannot be run.
Background removal, upscaling, and layer separation are distinct tasks, not features of every image model. Topaz Image Upscaler takes a source image and a scale setting without a prompt field. Layer decomposition takes one source with optional instructions; review its separate results in an external editor rather than expecting a full editing canvas in Studio.
Submit a first image task
Define the deliverable and review criteria. For a product listing, decide the placement, crop, background, and product details that must remain accurate before choosing a model.
Choose a compatible model. A text-only model cannot use an uploaded image as a visual anchor; an image-required operation cannot start with only a written description.
Write a focused prompt where the operation requires one. Example: “Create a concept image of a matte black bottle, centered on a light gray background, with space above for a heading and no props.” This is an illustrative prompt, not a tested output or a way to reproduce a real product exactly.
Upload references if supported or required. Wait for the upload to finish and check the attached previews, not just the files selected in the file picker.
Review the visible settings and credit estimate, then submit. Submitting creates a task; it does not mean the image is already ready. Check task lifecycle for status meanings.
Review the finished image against your criteria and save the accepted output. If you need another attempt, change one relevant part of the brief or settings so you can assess the difference.
For planning several related images, use image workflows. For a clearer written request, use prompt writing basics.
Use references and settings deliberately
State what each reference should guide: product appearance, composition, lighting, or style. A reference can guide generation; it does not guarantee exact geometry, text, logos, or brand consistency. Keep the original product file for comparison. See image references and safe upload guidelines.
File count, format, size, and dimensions vary by model. Do not apply one model's limits to the whole workspace. The file picker can accept only a subset when you select more files than the remaining slots. Verify which images appear after upload and follow references and uploads if one is rejected. Removing a preview removes it from the current form; it is not confirmation that an uploaded file was deleted from storage.
Aspect ratio, resolution, quality, output format, seed, and negative instructions are available only where the selected model exposes them. A seed is not a guarantee of identical images across models. Writing “4K” in a prompt does not replace a supported resolution setting. Recheck the credit estimate after changing parameters or references; some models charge for reference inputs as well as output.
Review results before using them
Inspect product shape, color, visible wording, logos, edges, reflections, shadows, and the intended crop. Check the actual output at the size you will use; a small preview can hide errors. Do not publish invented product features or claims simply because the image looks plausible.
When a task returns several images, select each preview to inspect it. Multiple results may be variants or outputs of a specific operation, such as separated layers; do not assume they form a finished composition. Background removal and upscaling also need review for lost edges, halos, and altered detail. They do not prove product accuracy or recover verifiable facts that were absent from the source.
Use History to find submitted tasks and the download guide to keep the files you need. Retain approved prompts, model and settings notes, references, and final files in your own project storage. Task history is not a permanent guarantee that every source or output file remains accessible.
Continue with the right image and budget
Each new generation or dedicated image operation is a separate task with its own cost. Check credits and billing; an estimate or reserved amount is not always the final settled charge. A completed image you dislike is not automatically a failed task eligible for a credit return. See failed tasks and refunds.
Switching models resets model parameters and can remove unsupported or excess references. A route change may carry only one compatible image and shorten a long prompt. Save your full brief first and review the new form before submitting; it does not preserve all settings as a project migration.
If an image result offers a shortcut to discuss it in chat, check the attachment there. This shortcut uses the task's primary result, which may differ from the variant currently selected in the preview. It opens a chat draft without sending the request. Choose a chat model that supports images before asking for visual feedback; see Chat Studio.
For motion, take the accepted image into a compatible Video Studio workflow and check that destination's input limits. Image generation does not itself animate the result or assemble a video. For narration or music, use Audio Studio separately.
Resolve errors before repeating a request
If upload fails, check the selected model's format, size, count, and dimension rules. If the browser disconnects after submission, check the task in History before submitting again; a disconnected page does not prove a failed task. Preserve unsent text before refreshing.
For an unresolved error, use troubleshooting with the model, task ID if present, approximate time, and visible error. If the task succeeded but the image is unsuitable, revise the prompt or references based on the specific defect instead of treating another run as a guaranteed fix.