by ElevenLabs
ElevenLabs' expressive text-to-speech model with inline audio tags, 90+ languages and Professional Voice Clone support.
Updated EU status checked Last shown
Sources: the vendor's own pages (linked above) and creator videos at the linked timestamps. EU status is information, not legal advice; see how we check.
Eleven v4 is ElevenLabs' text-to-speech model, described by the vendor as its most emotive voice model. It is built on a new architecture that interprets a script in context, including who is speaking and what has just happened. It supports speech in 90+ languages, multiple speakers, and built-in sound effects.
Direction is written into the script with inline tags such as [laughs], [whispers] and [door slams], and the vendor says v4 follows tag sequences more reliably than v3. Other listed capabilities are context stitching for long-form audio such as audiobooks, stable speaker identity when regenerating a line, and Professional Voice Clone support, which v3 lacked. Voices can come from a library of 20,000+ voices, voice design from a description, or cloning from about ten seconds of audio (instant) or a trained Professional Voice Clone. IPA support allows correcting pronunciation of specific words.
Eleven v4 Turbo is a companion low-latency variant for real-time use and agents, available through the API and ElevenAgents. The vendor cites ~100 ms median inference latency and ~150 ms median time to first speech, with bidirectional streaming. Both models are available in the ElevenLabs apps, API and agents. Access is via the hosted service only; weights are not published. Creators in coverage reported close voice clones and strong expressiveness.
Watch the moment: the exact second top creators show it. Videos load only when you press play, via youtube-nocookie.com.
ElevenLabs' latest text-to-speech model, described as the most emotive, with metatags to control direction, sound effects and tone. It is on the free plan and tops a TTS leaderboard.
ElevenLabs' most emotive voice model, steerable via style directions, with multiple languages and improved voice cloning. Creator played samples and found the clone very close.
1.7T-parameter MoE open model with D-Spark speculative decoding. Competitive with leading open models and some closed models; very cheap at about 6 cents per task on Artificial Analysis. Raw size 893 GB.
The latest DeepSeek V4 model with the DSpark mechanism already implemented, downloadable from Hugging Face. The creator says it should run faster and more efficiently than the previous version, which is already fast and cheap.
Open-weight text-to-image model with strong world knowledge, text rendering and prompt adherence, plus a bounding-box canvas workflow for controlling layout. Runs in ComfyUI (about 1 min per image, as low as 6GB VRAM with offloading). The creator calls it the best open-source image model, better than Z Image, Flux Klein and Qwen Image. It has a non-commercial license and is text-to-image only.
Image model with open weights, strong in typography and posters, using JSON prompts with bounding-box layout and layered outputs, including transparent images. ComfyUI workflow available; heavily censored; top open model on Arena, ranked ninth overall.
Closest neighbours by meaning, not just by category.
API · ElevenLabs
AI audio platform and API for text-to-speech, speech-to-text, music, sound effects, voice cloning and voice agents.
App · ElevenLabs
ElevenLabs' AI dubbing model that translates video and audio while carrying over the original speaker's voice and emotion.
App · ElevenLabs
ElevenLabs platform for building, deploying and monitoring real-time voice and chat AI agents.
Model · ElevenLabs
ElevenLabs' text-to-music model for generating songs with vocals, inpainting and genre transitions, cleared for commercial use.
Model · ElevenLabs
ElevenLabs' text-to-music model that generates tracks with or without vocals from natural-language prompts.
Model · Resemble AI
Open-weights expressive text-to-speech model from Resemble AI with voice cloning, fine-tuned from LTX-2.3 audio.
