Not named on pricing
The new pika.art pricing page doesn't mention Pikaformance or lip sync, and lists no per-second credit cost for it.
Pikaformance is Pika's audio-driven performance model. Give it a still image and an audio clip and it animates the face with lip sync, expressions and head movement, so your image can sing, speak, rap or even bark in time with the sound. This page covers what it does, how to use it, what it costs, and where Pika's talking-video tech is heading with PikaStream.
Pika introduced Pikaformance on the web with a simple pitch: hyper-real expressions synced to any sound, with near real-time generation speed. It grew out of Pika's earlier Lip Sync feature, which was built with ElevenLabs. Pikaformance animates the whole face, not just the mouth.
| Spec | Pikaformance |
|---|---|
| Input | A portrait or character image plus an audio track (speech, song, rap, sound) |
| Output | A video of the image performing the audio, with lip sync, expressions and head motion |
| Audio length | Up to 30 seconds (as listed on Pika's previous pricing) |
| Resolution | 720p (as reported) |
| Credit cost | 3 credits per second of audio on Pika's previous plans, so 30 credits for 10 seconds |
| Speed | Pika calls it its "first-generation ultra-fast video generation model": 8 GPUs and about 4.5 seconds of latency |
| Styles | Real photos, 3D, anime, cartoon and illustrated characters |
Speed figures from Pika's PikaStream announcement. Credit, length and resolution details from Pika's previous pricing, as reported by Sacra and RuntimeWire.
Pika rebuilt pika.art as a new platform of apps and models, and the talking-video options changed with it.
The new pika.art pricing page doesn't mention Pikaformance or lip sync, and lists no per-second credit cost for it.
The new site's lip-sync app turns a photo plus audio, or a stock avatar plus a script, into a talking clip. Pika doesn't say which model powers it.
The Pika app on iPhone still advertises "AI Templates, Effects, Lipsync" for turning selfies into talking or singing videos.
Pika's next-generation real-time face model, built for live video calls with AI agents. It runs on 1 GPU instead of Pikaformance's 8.
From Pika's pricing page, the Talking Head Video app and the Pika iOS app listing. If you're on old.pika.art or a legacy plan, Pikaformance may still appear there. Check your account.
On the new pika.art that's the Talking Head Video app. In the Pika iOS app, use a lipsync template.
Upload a clear, front-facing photo, or pick a stock avatar.
Upload your audio, or paste a script and pick a voice.
The app shows the credit cost before you generate; it depends on length and settings.
Watch the lip sync and expressions closely, then download. No Pika plan adds a watermark.
Steps based on Pika's Talking Head Video app page.
A clear, well-lit face looking at the camera gives the model the most to work with. Avoid hands, hair or glasses covering the mouth.
One voice, little background noise or music. Separate vocals before lip-syncing a song.
Shorter clips are cheaper to retry and easier to review. Split long scripts into sections.
A cartoon dog with a deep trailer voice is funny; a mismatch you didn't intend just looks off.
Pika says Pikaformance was fast for video generation but too slow for real-time conversation. In Pika's words, at 4.5 seconds every exchange felt like leaving a voicemail. PikaStream1.0 was built to fix that for live AI agents.
| Spec | Pikaformance | PikaStream1.0 |
|---|---|---|
| GPUs | 8 | 1 (H100) |
| Latency | About 4.5 s | About 1.5 s, speech to video |
| Output | Generated clips | Continuous stream, 24 FPS at 480p |
| Model | Not disclosed | 9B diffusion transformer, distilled to stream |
| Use | Talking and singing clips | Live video calls with AI agents |
Specs and videos © Pika, from the PikaStream1.0 announcement.
| Option | Starts from | Best for |
|---|---|---|
| Pikaformance / Talking Head Video | Image + audio or script | Making a still photo or character talk or sing |
| Pika H3 Flex | A prompt (+ references) | New scenes where characters speak, with dialogue generated in the video |
| Pika Speech | Text | The voice track itself, ready for lip sync or voiceover |
| PikaStream | Live speech | Real-time video calls with an AI agent |

More: H3 Flex audio guide, Pika Speech and Pika lip sync.
Only animate a real person's photo with their permission, and never put words in someone's mouth that they didn't say.
Use voices and songs you have the rights to. Cloned voices and commercial music need permission.
Turn on the AI-content label on YouTube, TikTok and Meta when a talking clip looks real.
See also: Pika watermark and AI labels.