Pika adds Lip Sync to give AI characters a real voice

If you have spent any time experimenting with AI video tools, you know the unspoken joke: characters either look mysteriously stoic or never speak at all. The moment they open their mouth, the illusion falls apart into an uncanny valley puppet show.

AI video startup Pika is aiming to change that. On February 27, the company rolled out a brand-new Lip Sync feature for its paying subscribers, teaming up with generative audio platform ElevenLabs to bring native character speech and synchronized mouth animations to AI videos.

With ElevenLabs under the hood, the feature handles both text-to-speech and custom audio uploads. That means creators can type out dialogue, pick a voice style, or record their own voice track, and Pika will animate the character's lips to match the audio naturally.

Access is currently restricted to early access users. You need to be a Pika Pro subscriber, which costs $58 a month billed annually upfront at $696, or hold an invite to Pika's exclusive Super Collaborators group on Discord.

Solving narrative film's biggest headache

Until now, putting dialogue into an AI video character meant relying on cumbersome third-party post-production tools. The results almost always had that disjointed, cut-and-paste feel reminiscent of classic Monty Python sketches.

By baking lip sync directly into the workflow, Pika removes one of the final stumbling blocks preventing creators from making full-blown narrative videos. While its raw visual fidelity may still lag behind OpenAI's headline-grabbing Sora and even Runway's latest outputs, Pika just gave creators a practical tool neither competitor currently offers natively.

Runway is not standing still, either. That same week, it upgraded its Multi Motion Brush with automatic region detection, letting users apply movement to distinct objects without manually painting over them. Pika still lacks that level of brush-based motion precision, even though it lets users expand canvases and edit parts of an image.

Not everyone in the creative world is celebrating, though. Ed Newton-Rex, founder of Fairly Trained and former VP of Audio at Stability AI, immediately took to X to press Pika on what training data was used for its video model. Meanwhile, Hollywood mogul Tyler Perry announced he is pausing an $800 million expansion of his production studio after seeing Sora, citing massive concerns over industry job losses.

Why it matters

We think Pika made a very smart tactical move here. While OpenAI keeps dazzling everyone with Sora demos behind closed doors, Pika is shipping tangible utility that working creators can actually put to use today.

For solo creators and digital marketing teams, this dramatically simplifies the pipeline for story-driven shorts, animated series, and talking-head explainers. You no longer have to bounce between three separate apps just to get a character to deliver a punchline. The real barrier right now is price: at nearly $700 upfront per year, casual hobbyists will likely wait for cheaper tiers to open up.

On an industry level, this launch shifts the AI video arms race from raw rendering quality to full multimodal production workflows. Cinema is audio and video combined, and Pika wisely avoided reinventing the wheel by partnering directly with ElevenLabs. The obvious risk, of course, is that Pika relies heavily on an external provider for voice, and scrutiny over video training data is only going to intensify. But for now, the era of silent AI video is officially winding down.