The AI voice cloning company is carving scripts read by a synced digital face out of its broader creative suite and giving them a dedicated product called Avatars in ElevenCreative.
ElevenLabs, the AI voice-cloning company, has carved talking-head video out of its broader creative suite and given it a dedicated product surface, Avatars in ElevenCreative. The bet is that the hard problems in this format, keeping a synthesized face in time with a synthesized voice across clips and languages, are not the same as the problems a general audio or video tool has to solve.
The new product combines text-to-speech, avatar generation, and lip-syncing in a single pipeline, according to the Product Hunt launch listing. A user writes a script, the system generates a voice, renders an avatar, and produces a video where the mouth moves with the audio. Avatars can be built from a photo or a text prompt and reused across clips, which the listing pitches as a way to keep an on-camera identity consistent across multiple videos. Multilingual generation is also advertised, with the same face intended to read in different languages.
That bundling is the interesting design choice. Most teams that want an AI presenter today stitch together a separate text-to-speech tool, an avatar generator, and a lip-sync post-process, then spend engineering time gluing the pieces together and cleaning up the seams. ElevenLabs is collapsing those steps into a single workflow and tying the whole thing to its Flows automation, which the Product Hunt listing describes as a way to script batch avatar videos. The implied bet is that the seams between voice, face, and timing are the actual product problem, not a side effect of using general tools.
The seams are also where skeptical practitioners land. One commenter on the Product Hunt listing, identified as having shipped async video for customer communications, called lip-sync accuracy the consistent bottleneck and asked how ElevenLabs handles sync drift over longer clips. The question is real. In current avatar pipelines, small audio-visual timing errors compound over a few minutes of footage, and mouth movements can lose lock with the audio in ways that are obvious to a viewer but hard to diagnose in QA. The launch listing does not address that, and ElevenLabs has not, in the public material reviewed for this piece, published technical details on how it handles accumulated timing error on takes longer than a sentence or two.
Pricing, safety and consent posture, and the specifics of availability beyond the Product Hunt launch page are not in the source. ElevenLabs did not have a primary blog post, release notes, or engineering write-up for the product in the materials reviewed here, so any market, adoption, or competitive framing has to wait for a second look at the company's own documentation and at least one independent third-party test.
What to watch: whether the company publishes lip-sync benchmark data on clips longer than a minute, and whether the avatar-to-voice pipeline is exposed as an API or stays locked inside ElevenCreative. Those two answers will tell readers whether the workflow split is a packaging decision or a real architectural bet on a different class of problem.