ElevenLabs releases SFX v3 sound effects model
ElevenLabs has released SFX v3, a sound effects model it says follows prompts that describe several sounds more reliably, a common problem for creators building full audio scenes from text.
ElevenLabs announced SFX v3, the latest version of its text-to-sound-effects model. It is available now in the company's ElevenCreative product and to developers through ElevenAPI.
The company names three main improvements over SFX v2. First, the model is better at following several instructions in a single prompt, so more of the sounds a user describes end up in the generated audio. ElevenLabs uses a football clip as an example: a crowd cheering, a ball dropping, and the same crowd booing seconds later. Until now, the company says, getting all of that from one prompt "took trial and error."
Second, SFX v3 handles prompts that include music and vocals more convincingly. That makes it possible to pair a musical phrase or vocal performance with environmental sounds.
Third, ElevenLabs says the model produces cleaner, more natural audio and is less likely to add harsh or crackly artifacts to sustained sounds such as gongs, chimes, and whistles.
ElevenLabs says the quality gains over SFX v2 come from "benchmark testing," but the post does not name the benchmarks or share any scores. It also does not cover pricing, latency, output length, or changes to the model's architecture or training.
In ElevenCreative, the model can be used in Flows, the Sound Effects tab, Studio, and the Music Editor. It is also in the ElevenLabs mobile app. Developers can add it to their own products through ElevenAPI.