I opened elevenlabs.io to map how speech, agents, music, and the API fit together — not to replace their docs, but to leave a readable field guide before you sign up.
Start with the job
Official use-case pages span video, audio, games, and developer embeds. Here are the five I keep pinned when someone asks “what is this for?”
Expressive long-form voices for chapters, characters, and localization.
Host voices, clones for pickups, and music beds without a full studio day.
Cast NPCs and trailers with distinct voices instead of booking every line.
TTS for YouTube, TikTok, ads, and campaign localization at draft speed.
Omnichannel agents that talk, type, and follow workflows with guardrails.
Capability showcase
Confirmed domains from the official site: speech and audio tools, music, transcription, agent dialogs, plus Creative image/video tooling. Cards below are local files or honest snippets — no filler covers.
Prompt notes for Eleven v3 (studio pass) [whispers] Even the birds fell silent when he passed. [sarcastically] Not the “burn it all down” kind… [giggles] but he was gentle, wise, with eyes like old stars. Takeaway: audio tags steer delivery mid-line without re-cloning the voice.
Eleven v3 audio tags — mid-line delivery control.
ElevenAgents sample flow (customer chat) User: Can I get a refund? Agent: Sure. Can you share your order number please? User: It's EL4543490 Agent: Thank you. I have initiated the order refund process. System: Refund completed
ElevenAgents refund dialog sketch from official demo pattern.
Scribe v2 sketch — meeting clip 00:00:01 Speaker 1: Ship the Spanish dub after legal clears the VO. 00:00:07 Speaker 2: Keep the laugh on the second take; cut the breath before the CTA. 00:00:14 Speaker 1: Timestamps to the word — editors will snap cuts to these.
Scribe v2 style transcript with speakers and timestamps.
Eleven Music v2 composition plan (editorial note) Prompt: Fast-paced electronic track for a product reel, bright vocals, 10s hook. Structure: intro → hook → short breakdown. Note: commercial rights follow the active subscription tier on the official site.
Music v2 composition-plan note for a short reel hook.
import { ElevenLabsClient } from "@elevenlabs/elevenlabs-js";
const client = new ElevenLabsClient({ apiKey: "YOUR_API_KEY" });
await client.textToSpeech.convert("JBFqnCBsd6RMkjVDRZzb", {
outputFormat: "mp3_44100_128",
text: "The first move is what sets everything in motion.",
modelId: "eleven_multilingual_v2",
});
Official-shaped Multilingual v2 TTS call.
import { ElevenLabsClient } from "@elevenlabs/elevenlabs-js";
const { music } = new ElevenLabsClient();
const compositionPlan = await music.compositionPlan.create({
prompt: "Fast-paced electronic track for a video...",
musicLengthMs: 10000,
});
Music API compositionPlan create — from docs shape.
Platforms
The homepage frames ElevenCreative, ElevenAgents, and ElevenAPI as sibling platforms. I keep them separate so a podcast brief does not get confused with a contact-center rollout.
About
ElevenLabs builds foundational audio models and the products around them. On the pages I opened, the company describes a path that started with human-like voice and now spans transcription, music, voice cloning, dubbing, intelligent agents, and a Creative suite that also reaches into image and video generation with partner models.
Day to day I treat it as a production audio stack: write a line, pick a voice from a large library or a clone, generate speech with Multilingual v2 or Eleven v3, then — when the brief needs it — add Music v2 beds or Scribe transcripts. Agents sit on a different track: configure behavior, test conversations, watch resolution metrics.
Enterprise logos on the homepage (from Twilio to Epic Games) signal who already ships on the platform. I still verify features against the current models table before promising latency or language counts to a client.
For the product itself, go to the official ElevenLabs website (elevenlabs.io). This site stays an editorial map with local next-step buttons.
Open the local next step, or jump into the model catalog if you already know whether you need Flash latency or v3 expression.
Benefits
Eleven v3 audio tags let me shape whispers, sarcasm, and laughter mid-line without re-casting the voice.
Flash v2.5 is positioned for ~75ms conversational paths; Conversational v3 trades some speed for richer delivery.
Creative, Agents, and API share the model line — I can draft in the editor and ship the same voice via API.
Scribe v2 and Realtime close the loop from spoken notes back to editable text with diarization.
Music v2 prompts and section edits cover short hooks; commercial rights follow the official plan tier.
Testing, analytics, guardrails, and workflows are first-class on ElevenAgents — not an afterthought slide.
Models
Names below match the official models docs. Deprecated Turbo entries are omitted in favor of Flash.
Most expressive speech model: dramatic delivery, audio tags, 70+ languages, multi-speaker dialogue.
Realtime expressive speech for agents and interactive characters (~280ms model latency).
Lifelike, consistent TTS across 29 languages — strong for long-form narration stability.
Ultra-low latency TTS (~75ms) for conversational apps; 32 languages; leaner per-character cost.
Speech recognition with word-level timestamps, diarization, entity detection, 90+ languages.
Streaming transcription with low latency (~150ms) for live captions and agent loops.
Studio-grade music from natural-language prompts — vocals or instrumental, editable sections.
These notes are independent. Product truth lives on the official ElevenLabs website. Primary buttons on this page open a local dialog, not a login.
Who it fits
YouTube, podcasts, and ads that need fast VO and optional music beds inside ElevenCreative.
NPC lines, trailers, and localization without booking every language take.
Teams piloting voice agents with testing, guardrails, and omnichannel routes.
Anyone wiring TTS, STT, music, or agents through ElevenAPI with the JS or Python SDKs.
How to start
Creative for media, Agents for CX, API if you already have a client. Do not mix the brief.
v3 for expressive VO, Flash for realtime, Scribe for transcripts, Music v2 for beds.
Check language counts, latency notes, and plan limits on the official site before you promise a ship date.
FAQ
ElevenLabs is an AI audio company with three product surfaces I keep straight: ElevenCreative for making speech, music, sound effects, images, and video; ElevenAgents for deployable voice and chat agents; and ElevenAPI for the same research stack in code. This page is an independent overview. It is not the official ElevenLabs site. For accounts and downloads, use the official ElevenLabs website.
The homepage I resolved is the official ElevenLabs website at elevenlabs.io. Product docs for models live under their docs overview. Press and brand assets are on the official press and brand pages. I treat those as the source of truth, not social recaps.
If you need expressive, tagged delivery and can wait for studio quality, start with Eleven v3. If the app talks back in realtime, Flash v2.5 is the low-latency path the docs emphasize (~75ms model time). Multilingual v2 remains the steady long-form option across 29 languages. Conversational v3 sits between expressiveness and agent latency (~280ms).
No. The official homepage and Creative platform also cover music, sound effects, dubbing, speech-to-text (Scribe), and — inside ElevenCreative — image and video tools that call out models such as Veo, Wan, Kling, and Seedance. This overview still centers audio because that is where the foundational models and most of my notes sit.
Official copy positions ElevenAgents for omnichannel customer experiences: phone, chat, email, WhatsApp, with analytics, testing, guardrails, and workflows. Expressive Mode is listed for more natural agent delivery. I treat agents as a product line, not a side demo of TTS.
No. This is an independent notes page. Product facts come from pages I opened on elevenlabs.io and the models docs. Primary buttons here open a local next-step dialog. To use the product, follow the text link to the official ElevenLabs website.
Five reader impressions for this overview, not official product reviews.
Maya K.
5 / 5
I rebuilt a product demo VO in Eleven v3 with audio tags. The whisper-to-punch line landed on the first keep. Editors stopped asking for another booth session.
Owen R.
4 / 5
Flash v2.5 is what I wire into the agent stack. Multilingual v2 still wins for the documentary chapter reads. I keep both keys in the same project.
Sofia L.
5 / 5
Scribe v2’s diarization cleaned a messy panel recording. Word timestamps made the cut list boring — in a good way.
Kenji T.
5 / 5
Music v2 gave me a 10-second hook that matched the brand reel. I edited one section and shipped. Rights still follow the plan on the official site.
Aisha N.
4 / 5
ElevenAgents handled a refund flow we scripted in testing. Guardrails mattered more than the pretty voice. I still review transcripts weekly.
Use the local next step here, then move to the official product when you want an account. The catalog stays on this site if you need model pages first.