Stable Audio 2.0
Open Latent-Diffusion Audio
Stability AI's open text-to-audio generator, releasing 3-minute 44.1kHz stereo tracks from text prompts. Less viral than Suno (which had launched v3 two weeks earlier as a polished consumer product), but the only open-weight option in the category — a role analogous to what Stable Diffusion played in image generation while DALL-E was closed.
Cost
Free
Open weights — self-host
How are Intelligence, Speed & Cost bucketed?
Intelligence and Speed buckets are percentile ranks on
Artificial Analysis. Cost buckets are fixed dollar
thresholds keyed off output-token price ($/M out).
Intelligence
- Top 1%≤ 1%
- Top 5%≤ 5%
- Top 10%≤ 10%
- Good≤ 25%
- Medium≤ 50%
- Below avg> 50%
Speed
- Top 1%≥ 345 tok/s
- Top 5%≥ 237 tok/s
- Top 10%≥ 196 tok/s
- Good≥ 146 tok/s
- Medium≥ 90 tok/s
- Slow< 90 tok/s
Cost
- Freeopen weights · self-host
- Low< $1 / M out
- Moderate$1–5 / M out
- High≥ $5 / M out
Why it matters
Stable Audio 2 is the open baseline that makes the closed audio-AI category contestable. Without it, Suno / Udio / ElevenLabs would have unchallenged pricing power.
Core Capabilities
Audio
Speech, music, or other audio understanding/synthesis.
Generative
Produces images, video, audio, or other media.
Multimodal
Combines text, vision, and audio in one model.
Context Window
Context window not disclosed.
Availability
API
Not available
Product / App
Not available
Open Source
Released
Enterprise
—
Pricing Model
Free / self-host
Open weights — pay only for compute.
Self-host Capability / Performance
Where this model sits relative to the middle 60% of models in the tree. All scores are 0–10 (higher is better).
Lower 20% Upper 80% This model
Quality
No data reported · placeholder
5.0
Speed
No data reported · placeholder
5.0
Control
No data reported · placeholder
5.0
Consistency
No data reported · placeholder
5.0
Lower 20% 20th percentile — 20% of models score below this This model Where the current model lands Upper 80% 80th percentile — only 20% of models score above this
Percentile boundaries are computed across every model in the tree that reports the underlying benchmark for each capability.
What it feels like
- Stability's audio diffusion sibling — same lab heritage.
- First text-to-image model with DALL·E 2-class quality and permissive open weights
- Latent diffusion innovation — denoising in compressed latent space, not pixel space — made consumer-GPU inference viable
- Ran on <8GB VRAM at release — first generative model regular people could use locally
Best use cases
- Self-hosted image generation pipelines (privacy / volume / customisation)
- Custom-style fine-tuning via LoRA / textual inversion / Dreambooth
- ControlNet-style guided generation requiring weight access
Tools to try
Not ideal for
- Out-of-the-box photorealistic aesthetics — Midjourney still the default for that
- Reliable in-image text rendering (FLUX.2 and later models leapfrogged here)