Generation Guide

How generation works

Every generation on Rhyora follows a simple flow:

1. Choose a studio (Image, Video, Audio, Music, 3D, or Avatar).
2. Hit Generate — AI automatically selects the best model and fills in optimal settings based on your shot context. Credits are pre-held and the job enters the queue.
3. Watch the real-time progress indicator as your asset is created, with elapsed time displayed.
4. Download, iterate, or approve the result.

Want to choose a specific model or adjust settings? Click “Change model / Advanced” to access the full model library and all parameters.

If a job fails for any reason, your credits are automatically refunded in full.

Automatic model selection

When working inside a project, Rhyora automatically picks the best model for each shot based on your creative context — the references you've linked, the shot type, camera movement, and visual style. The AI selects from three pricing tiers:

• Budget — Fast and affordable. Ideal for drafting, iteration, and exploring ideas quickly.
• Value — Balanced quality and cost. Best for most production work.
• Premium — Highest quality output. Use for hero assets, final renders, and client-facing work.

You'll see which model was selected and can switch to a different one at any time by clicking “Change model / Advanced”. For standalone generations outside of projects, you choose the model yourself.

Image Studio

The Image Studio is the most versatile workspace, supporting two primary modes:

• Text-to-image — Describe what you want in a text prompt. Be specific about style, composition, lighting, mood, and subject matter for the best results.
• Image-to-image — Provide a reference image along with a prompt. The AI uses the reference as a starting point and transforms it according to your instructions. Adjust the strength parameter to control how much the output deviates from the input.

Common parameters include aspect ratio (portrait, landscape, square, and custom), quality level, seed (for reproducible results), and negative prompts (to exclude unwanted elements). Not all parameters are available on every model — the form dynamically adapts to each model's capabilities.

Video Studio

The Video Studio supports multiple generation modes for different creative needs:

• Text-to-video — Generate a video clip directly from a text description. Best for quick concepts and motion studies.
• Image-to-video — Animate an approved still frame into a video. This is the recommended workflow for production-quality output: generate and approve a frame in Image Studio first, then bring it to Video Studio.
• Video extend — Extend an existing video clip with additional frames, continuing the motion and narrative.
• Video transform — Apply style changes to an existing video while preserving its motion and composition.
• Visual effects — Add VFX treatments, transitions, and stylistic effects to video clips.
• Character animation — Animate characters with motion, expressions, and gestures from text or audio input.

Important: For image-to-video, the source image must be approved first. This ensures you're investing video credits in a frame you're happy with. Approve any image from its detail view or the review panel.

Duration options vary by model. Each model supports specific clip lengths:

• Veo 3.1 — 4, 6, or 8 seconds
• Kling 2.6 Pro — 5 or 10 seconds
• Hailuo 2.3 — 6 or 10 seconds
• WAN 2.6 — 5, 10, or 15 seconds

When working within a project, shot durations are pre-planned using these values. The duration you select at generation time should match the planned shot duration for consistent pacing across your production.

Audio Studio

Generate spoken audio and sound effects with AI:

• Text-to-speech — Convert written text into natural-sounding speech. Choose from multiple voices with different accents, genders, and tonal qualities. Adjust speed and emphasis for the delivery you need.
• Sound effects — Describe an ambient sound, foley effect, or audio texture and the AI generates it. Useful for adding atmosphere to video projects.

Output formats include MP3 and WAV. Audio assets integrate directly into your project timeline alongside visual assets.

Music Studio

Create original music tracks and scores:

• Text-to-music — Describe the mood, genre, tempo, and instrumentation you want. The AI generates an original instrumental track.
• Audio-to-music — Provide a reference audio clip and the AI generates music that matches its style, tempo, or mood.

Music assets are great for scoring video projects, creating background tracks for podcasts, or generating royalty-free music for content.

3D Studio

Create 3D objects from text descriptions or reference images:

• Text-to-3D — Describe an object and the AI generates a textured 3D model.
• Image-to-3D — Provide a reference image and the AI infers the 3D structure and generates a model that matches it.

Output formats include GLB and OBJ, ready for import into game engines (Unity, Unreal), 3D editors (Blender), AR/VR applications, and web viewers.

Avatar Studio

Create character animations and talking-head videos:

• Lip-sync — Provide a portrait image and an audio track. The AI animates the face to match the speech, producing a realistic talking-head video.
• Face animation — Animate facial expressions and head movements on a portrait image from a reference video or audio input.
• Face swap — Transfer one face onto another in a video or image, maintaining the original motion and expressions.
• Portrait transfer — Apply the style or likeness of a portrait across different contexts and poses.

Avatar models work best with clear, front-facing reference images and high-quality audio input for lip-sync.

Pipeline progress

When generating within a project, a progress strip shows the production pipeline for each shot: Image → Video → Audio. Completed stages are marked with a checkmark, and you can click any stage to switch to that workspace. Each active generation shows elapsed time so you always know how long it's been running.

Automatic prompt optimization

When using the Rhyora Prism (Guided or Auto mode), your shot prompts are automatically enhanced before the first generation attempt. The optimizer adds technical photography terms, lighting specifics, and composition guidance that image models respond to best.

This happens behind the scenes — you don't need to write perfect prompts yourself. The original prompt is preserved; the optimized version is used only for generation. If optimization fails for any reason, the original prompt is used as-is so production is never blocked.

Quality evaluation

After each image is generated through the Rhyora Prism, an AI vision model evaluates the result against your original prompt. It scores four dimensions:

• Prompt adherence — Does the image match what was described?
• Composition — Is the framing, balance, and visual flow strong?
• Technical quality — Sharpness, lighting, color accuracy.
• Mood — Does the emotional tone match the scene context?

The combined score (1-10) is compared against your quality threshold. Images that pass are approved. Images that don't are retried with a refined prompt that addresses the specific issues found. Each shot gets up to two quality retries before auto-approving to keep production moving.

Prompting tips

The quality of your output depends heavily on your prompt. Here are tips that apply across all studios:

• Be specific — Instead of “a cat,” try “a tabby cat sitting on a windowsill, afternoon sunlight, soft bokeh background, photorealistic.”
• Describe style and mood — Include art style (cinematic, anime, watercolor), lighting (golden hour, studio lighting, neon), and mood (serene, dramatic, playful).
• Use negative prompts — Where supported, negative prompts exclude unwanted elements (e.g. “blurry, low quality, text, watermark”).
• Iterate with budget models — Refine your prompt using fast, low-cost models. Once you're happy with the composition and style, switch to a premium model for the final render.
• Use the prompt optimizer — The built-in AI prompt optimizer rewrites your prompt to be more detailed and model-friendly. It adds technical parameters that improve output quality.
• Use seeds for consistency — If you find a result you like, note its seed value. Using the same seed with the same prompt produces similar output, letting you make small prompt tweaks while keeping the overall composition.

Parallel generation

You can submit multiple generation jobs at once. Each plan tier includes a concurrency limit — the number of jobs that can run simultaneously:

• Free — 4 parallel tasks
• Starter — 6 parallel tasks
• Creator & Pro — 8 parallel tasks
• Studio — 12 parallel tasks

Jobs from different studios can run at the same time — generate an image while a video renders, or produce audio while a 3D model is being created.

On the storyboard, the Produce All button shows how many parallel slots are available so you know how many shots will start immediately. When you reach the limit, remaining shots start as slots free up. To increase your parallel limit, upgrade your plan.

Frequently asked questions