Eleven Labs is split across two provider keys

KIE holds Multilingual and Speech Turbo, Eleven Labs holds Voice Change, so neither key alone is the brand roster, and KIE is never a hub.

The catalog's provider column is an ingest label, not a brand. Eleven Labs is the case that makes that sentence load-bearing. Search the raw key "Eleven Labs" and you get Voice Change. Search "KIE" and you get Multilingual and Speech Turbo. Add the two keys and you have the brand's three models. Treat either key as the roster and you are looking at a fragment.

Two keys, one brand

Providers groups hubs by who actually builds the model, not by the ingest string that happened to land on the row. Three catalog quirks would produce wrong pages if used raw. One of them is this split: Eleven Labs is filed across two keys, so neither key knows the brand has three models.

The rows:

• Multilingual — text-to-speech, filed under KIE. The published spec is ElevenLabs Multilingual.
• Speech Turbo — text-to-speech, also filed under KIE. Fast single-language generation. It is a catalog row, not a second hub.
• Voice Change — audio-to-audio, filed under the "Eleven Labs" key. The published spec is ElevenLabs Voice Change.

KIE is an API aggregator key, not a brand a user searches for. The two models filed under it are Eleven Labs models and are attributed as such when the brand map is applied. The "Eleven Labs" key, on its own, holds Voice Change. A hub titled from either key would be a list of one job pretending to be the company.

Why KIE is never a hub

The provider index has a hard refuse list. "Other" is a catch-all bucket, not a brand — a /providers/other page would be a hub for "we did not classify this." "KIE" is the other refuse: an aggregator label. It cannot earn a roster page regardless of how many rows sit under it.

That is not a slight on the models. It is a classification rule. A hub's job is "show me everything this brand has." An aggregator key is not a brand. Publishing /providers/kie would teach the wrong noun.

The Eleven Labs key also does not earn a hub. A roster needs enough models and enough spec pages to be worth reading as a list. Voice Change alone is one row. One row is already the model page. A hub over it would be a thinner copy of Voice Change.

So there is no /providers/eleven-labs and there is no /providers/kie. The brand is visible as three engine rows on the speech surfaces, not as a provider card.

The three jobs, named separately

AI text to speech is the capability shelf: every model metered per 1,000 characters of script — the ones that read text aloud. Multilingual and Speech Turbo sit on that shelf. Voice Change does not. It is audio-to-audio: you supply speech, it transforms the voice, it keeps the words. Putting Voice Change on a TTS ranking would be the same error as putting a background remover on a video-generator ranking.

Voice-over is the language surface. Eleven Labs is one of the engines listed there, with three catalog slugs: Multilingual, Speech Turbo, Voice Change. The engine's job, in that page's terms, is multilingual voices, emotion tags and voice changing. ISO 639-1 is how the language is addressed. That is the brand roster in the place a user actually looks for a voice, without pretending KIE is a company.

Do not merge the three jobs because the logo is the same. A multilingual read, a turbo read, and a voice conversion are three meters and three inputs. Name the row.

Where to look instead of a missing roster page

If you arrived here from "Eleven Labs models list," the honest answer is: there is no provider hub, on purpose. Use:

1. The two published spec pages for the rows that have them.
2. The TTS tool for anything that starts as a script.
3. The voice-over language pages for named engines and what each engine adds.
4. The providers index for brands that did clear the roster gate — Kling, Google, Wan, Flux. Eleven Labs is not missing from that list because someone forgot it. It is missing because neither ingest key is a brand roster.

The same pattern shows up elsewhere in the catalog: Flux image and Black Forest Labs video are two keys on purpose; Alibaba's Wan, Qwen and Happy Horse lines keep separate hubs. Splitting keys is how the site avoids answering the wrong query. For Eleven Labs the split is messier — the keys do not even add up to a hub — so the attribution lives in the brand map and on the speech pages, not on a roster URL.

FAQ

Why isn't there a /providers/eleven-labs page?

Neither ingest key is a complete brand roster, and KIE is never allowed to become a hub. Voice Change alone does not clear the model-count and spec-page gate. The three rows are listed on the speech surfaces instead.

Is KIE an Eleven Labs competitor?

No. KIE is an API aggregator key in the catalog. The models filed under it here are Eleven Labs Multilingual and Speech Turbo. It is not a brand to search for and not a roster to browse.

Does Voice Change count as text-to-speech?

No. It is audio-to-audio voice conversion. The TTS tool is for models that read a script. Voice Change needs speech in.

Where do I pick an Eleven Labs voice for a language?

/voice-over is the language index. English, Spanish, Korean and French are the language pages that shipped. The Eleven Labs engine is listed there with the rest of the speech roster; it is not a second provider hub.