ElevenLabs v3 is the most expressive model in the ElevenLabs lineup. It reads inline audio tags such as [whispers], [excited], and [sighs] as performance direction, supports more than 70 languages, and renders multi-speaker dialogue as a single coherent scene.
In Phygital+ it is the default model in Sound Creation and Dialogue Creation, so speech, sound effects, and two-speaker scripts all run from the same canvas as your video and image nodes.
Voice-over, narration, character dialogue, ads, explainers, and localised audio
Content, localisation, and game teams that need delivery, not just intelligible speech
Inline performance direction, 70+ languages, and multi-speaker dialogue in one model.
Bracketed cues like [whispers], [excited], or [sighs] go straight into the script and tell the model how to deliver the line. It is closer to directing an actor than configuring a synthesiser, and it happens inside the text rather than in a settings panel.
More than 70 languages, up from 29 in Multilingual v2. A cloned voice keeps its accent character across all of them, so the same speaker stays recognisable in every localised version instead of turning into a different person per market.
Two speakers, two scripts, one generation. The model matches prosody across the exchange, so reactions and interruptions land the way a conversation actually does rather than sounding like two monologues cut together.
Chemical formulas, numbers, abbreviations, and other awkward strings that earlier models mangled are handled far more reliably. Less time spent respelling words phonetically to make them come out right.
Build a voice from a written brief with Voice Creation, or clone one from audio samples with Voice Clone, then reuse that Voice ID across every generation. Recurring characters and a consistent brand voice stay consistent across clips.
ElevenLabs v3 fits wherever the delivery matters as much as the words: narration, character dialogue, ads, and localised versions of the same script.
Write the script, place audio tags inline, pick a voice, and generate.
Add a Sound Creation node for narration and sound effects, or Dialogue Creation for a two-speaker scene. Both default to the eleven_v3 model.
Type the script and place audio tags inline where you want direction: [whispers], [excited], [sighs]. Punctuation controls breathing and pauses, so commas and full stops matter. Pick a preset voice, or paste a Voice ID from Voice Creation or Voice Clone.
Fix a seed to reproduce a take, and use Previous text and Next text to keep tone consistent across a series of clips. Chain the audio into a video node, or export it with commercial rights on your plan.
One subscription across 30+ AI models, with no per-tool credit balances or separate signups. Credit cost per generation is shown live in the node before you run it.
500 weekly credits to test the model and hear the output quality. No credit card required.
45,000 monthly credits with cheaper per-credit pricing, commercial rights, and full downloads.
90,000 monthly credits, up to 10 seats, shared workspace, centralized billing, and priority support.
210,000 monthly credits, API and integration support, unlimited seats, and dedicated management.
Full specs for ElevenLabs v3 in Phygital+.
How ElevenLabs v3 compares with the other ElevenLabs models available in Phygital+.
Chain scripting, voice design, speech, and video into one repeatable workflow.
Write the script with a text node, then send it straight into speech generation so wording and delivery are decided in one pass.
Build a voice with Voice Creation or clone one with Voice Clone, then reuse that Voice ID across every v3 generation.
Generate the visuals with a video node and drop the v3 track underneath for narration, dialogue, or sound design.
Generate ambience and effects in the same node, then layer them under the voice track for a finished scene.
Everything you need to know about ElevenLabs v3 in Phygital+.
Eleven v3 is ElevenLabs' third-generation text-to-speech model. It was released as a public alpha in June 2025 and reached general availability in 2026. Its defining feature is inline audio tags: bracketed cues written directly into the script that direct emotion, pacing, and non-verbal reactions.
Audio tags are bracketed instructions you place inside the text itself, such as [whispers], [excited], [sighs], or [laughs]. They tell the model how to deliver the line rather than what to say. Earlier ElevenLabs models handled basic pauses and breaks; v3 responds to a full range of emotion and delivery direction.
More than 70, up from 29 in Multilingual v2. A cloned voice keeps its accent character across languages, and the same synthetic voice stays recognisably the same speaker in each one, which matters when you are localising a series rather than a single clip.
Yes. In Phygital+, the Dialogue Creation node runs the v3 text-to-dialogue workflow: two speakers, separate scripts, a semicolon between phrases, and preset voices or custom Voice IDs per speaker. The model matches prosody across the exchange instead of stitching two independent monologues together.
It trades latency for fidelity. v3 uses a larger model and a higher-fidelity codec, so it takes longer to run and is not intended for live conversational agents. For real-time use, Flash v2.5 or Turbo v2.5 are the right choices, and both are available in the same Sound Creation node.
In Phygital+ you reach it through three nodes. Sound Creation covers text-to-speech and sound effects. Dialogue Creation covers two-speaker scenes. Voice Creation and Voice Clone build the voice identities that the first two use.
Punctuation carries prosody, so commas and full stops do real work. Add audio tags for direction, and use the Previous text and Next text fields when generating a series of clips so tone and pacing stay consistent across them. Results vary by voice, so it is worth testing tags on the specific voice you plan to ship.
Usage is credit-based and included in every plan, from Free up to Enterprise. The credit cost per generation is shown in the node before you run it, so there is no separate ElevenLabs subscription to manage.
Outputs created on any paid plan come with commercial-use rights. Check the model card for provider-specific limits, and make sure you have permission for any voice you clone from a real person.
Browse the full catalog of 30+ AI models available in Phygital+.
Join 100+ teams using Phygital+ – every model in one workspace
Try Phygital+ free
