- Text Generation
Claude Sonnet 4.5 is an Anthropic large language model optimized for software development, computer use, and agentic workflows, offering strong performance on coding and reasoning tasks…
Powered by Google
Lyria 3 Clip Preview
- Text Generation
Lyria 3 Clip Preview is Google's preview music-generation model optimized for creating short, 30‑second musical clips, loops, and previews from text or image prompts.
Providers Comparison
Speed
Output tokens per second · Higher is better
L
LLM API
S
SiliconFlow (FP8)
G
GMI
C
CoreWeave
N
Novita
D
DeepSeek
P
Parasail (FP8)
D
DeepInfra (FP4)
End-to-End Response Time
Seconds · Lower is better
L
LLM API
S
SiliconFlow (FP8)
G
GMI
C
CoreWeave
N
Novita
D
DeepSeek
P
Parasail (FP8)
D
DeepInfra (FP4)
Price
USD per 1M tokens (blended) · Lower is better
L
LLM API
S
SiliconFlow (FP8)
G
GMI
C
CoreWeave
N
Novita
D
DeepSeek
P
Parasail (FP8)
D
DeepInfra (FP4)
About the model
What is Lyria 3 Clip Preview?
Lyria 3 Clip Preview is a Google music-generation model that produces high-quality short audio clips from text and image inputs. It is mainly used to generate 30-second music snippets, loops, and previews for creative, media, and sound design workflows. It also supports features like vocal or instrumental modes, user- or model-generated lyrics, and controls such as BPM and intensity to shape the resulting clip. It belongs to the Lyria 3 family of music-generation models, alongside Lyria 3 Pro, and is offered as a preview model via the Gemini API and related Google Cloud platforms.
Model capabilities
5 Core Capabilities
-
Text-to-music
Generates 30-second high-quality stereo music clips from detailed text prompts, including structure, style, mood, and instrumentation guidance.
-
Image-to-music
Creates musical clips and previews conditioned on input images, translating visual themes and scenes into coherent audio compositions.
-
Vocal And Lyrics
Supports vocal generation, lyric generation, and user-provided lyrics to produce clips with synchronized singing and musical phrasing.
-
Musical Controls
Offers BPM and intensity controls plus instrumental mode, enabling tailored rhythmic feel, energy, and arrangement for generated clips.
-
Safety Filtering
Applies input filtering, output recitation filtering, and vocal similarity filtering, alongside audio watermarking for safer music outputs.
Use cases
6 Most Valuable Use Cases
- Short Music Clips
- Loop Background Tracks
- Ad Jingle Generation
- Social Media Audio
- Game Sound Previews
- Image-To-Music Promos
Transparent pricing
Cost Comparison
LLM API offers the lowest cost and latency for Lyria 3 Clip–class models.
| Provider | Region | Latency | Throughput | Uptime | Input ($/1M) | Output ($/1M) | Context |
|---|---|---|---|---|---|---|---|
| Global | ~550ms | ~8 img/s | 99.9% | ~$1.20/1K images | $0.00 | ~10 min video or 20 images | |
| Vertex AI (Google Cloud) | US East | ~480ms | ~10 img/s | 99.9% | ~$1.30/1K images | $0.00 | ~10 min video or 20 images |
| Replicate | US West | ~750ms | ~5 img/s | 99.5% | ~$1.80/1K images | $0.00 | ~8 min video or 16 images |
| LLM API BEST | Global | 180ms | 20 img/s | 99.99% | $0.80/1K images | $0.00 | 12 min video or 32 images |
Performance benchmarks
Technical Specifications
| Metric | Lyria 3 Clip Preview (Google) | CLIP ViT-L/14 (OpenAI) | SigLIP Large Patch16-384 (Google) |
|---|---|---|---|
| Avg Latency | ~800ms | ~700ms | ~900ms |
| Context Window | 128K | 128K | 200K |
| Input Price ($/1M tokens) | ~$0.20 | $0.15 | $0.25 |
| Output Price ($/1M tokens) | ~$0.60 | $0.60 | $1.25 |
| Max Output Tokens | 8K | 16K | 4K |
| Throughput | ~80 tps | ~100 tps | ~60 tps |
| Uptime | ~99.9% | ~99.9% | ~99.9% |
30-day usage via LLM API
- 620M
- Prompt tokens processed (30 days)
- 7.5M
- Completion tokens generated (30 days)
- 410K
- API requests served (30 days)
- 98.9%
- Avg uptime over last 30 days
Architecture & Integration
Why Build on LLM.API?
One unified API. Every major model. Built-in reliability, cost control, and observability.
-
Unified AI Routing
Dynamically route each request to the optimal model across providers based on latency, cost, and performance—without changing your code or integrations.
One endpoint, every model -
Cost-Aware Orchestration
Optimize spend by automatically choosing cheaper equivalents, downgrading when quality allows, and enforcing per-project budgets with centralized cost controls and analytics.
Max performance, minimal spend -
Resilient Fallback Logic
Eliminate single-vendor downtime with automatic failover to backup models and providers, using configurable rules, health checks, and graceful degradation strategies.
Stay online, even when APIs fail -
Deep LLM Observability
Trace every call across providers with logs, metrics, and structured events to debug prompts, track latency, and understand model behavior in production.
See every token, every hop -
Task-Level Abstractions
Describe tasks—chat, generation, extraction, tools—once and let LLM.API select and configure the right models, prompts, and parameters for each use case.
Think in tasks, not models -
High-Throughput Batch Jobs
Run large-scale batch inference with automatic chunking, concurrency control, retries, and progress tracking designed for data pipelines and offline processing.
Ship millions of calls safely
Decision guide
When to Use — When NOT to Use
Use it if...
- You need to generate short, engaging video clips from text or prompts.
- Your use case involves rapid prototyping of social media clips and story previews.
- You need an automated way to create video teasers for marketing or product launches.
- Your use case involves experimenting with AI-generated visual narratives or concept clips.
- You need a model to help non-experts quickly draft video content variations.
- Your use case involves embedding clip generation into creative tools or content pipelines.
Avoid if...
- You need advanced text understanding, reasoning, or general-purpose conversational intelligence.
- Your workload requires high-precision image classification or detailed computer vision analytics.
- You need long-duration, coherent video generation beyond very short clip segments.
- Your workload requires strict, fine-grained control over camera movements and visual continuity.
- You need domain-specific scientific visualization or technical diagram generation with guaranteed accuracy.
- Your workload requires on-device or fully offline inference without cloud dependencies.
FAQ
Frequently Asked Questions
-
What is Lyria 3 Clip Preview?
Lyria 3 Clip Preview is a Google model for generating short video clips from text prompts, accessible through the unified LLM.API gateway.
-
What is Lyria 3 Clip Preview best suited for?
It is best for quickly prototyping and previewing short video concepts, storyboards, and motion ideas directly from text descriptions.
-
What modalities does Lyria 3 Clip Preview support?
Lyria 3 Clip Preview consumes text prompts and outputs short video clips, without support for audio or image-only inputs in this preview tier.
-
How is Lyria 3 Clip Preview priced on LLM.API?
Pricing is usage-based per generated clip or generated video-second, with exact rates defined in your LLM.API pricing dashboard.
-
What is the context window or prompt size for Lyria 3 Clip Preview?
The model supports moderately long text prompts, typically a few paragraphs, but you should avoid excessively long scripts or scene-by-scene screenplays.
-
How fast is Lyria 3 Clip Preview in terms of latency?
Latency depends on clip length and load, but you should expect generation to take several seconds up to a couple of minutes per request.
-
How do I call Lyria 3 Clip Preview via LLM.API?
Use the standard LLM.API generation endpoint, specifying the Google provider and the "Lyria 3 Clip Preview" model name in your request payload.
-
How does Lyria 3 Clip Preview compare to other video models on LLM.API?
It emphasizes quick preview-quality clips for ideation rather than long-duration, high-fidelity cinematic video compared to heavier video generation models.
-
What are the main limitations of Lyria 3 Clip Preview?
Limitations include short clip duration, preview-level visual quality, occasional motion artifacts, and imperfect adherence to very detailed or complex scene instructions.
-
Can I use Lyria 3 Clip Preview for production-ready marketing videos?
It is better suited for concepting and iteration; production workflows typically require post-processing or higher-fidelity models for final output.
Compare
Competitive Models
- Text Generation
Devstral 2 2512 is a 123B-parameter open-source large language model from Mistral AI, optimized for agentic coding and long-context software engineering workflows. It supports a 256K/262K-token…
- Text Generation
Grok Imagine Image Quality is an image-focused evaluation or enhancement component from xAI’s Grok ecosystem, aimed at assessing or improving the visual fidelity of generated images.…