CPMV AI is the online alternative to Large Performance Model (LPM). Turn any text prompt into expressive character performance videos — the same LPM-quality character animation, available right in your browser. No waitlist, no download.
LPM (Large Performance Model) is a groundbreaking AI research project by miHoYo that generates realistic character performance videos — facial expressions, gestures, body language, and conversational reactions. LPM 1.0 introduced the concept of a Character Performance Model: an AI system that doesn't just animate a face, but creates a full performer who listens, speaks, reacts, and emotes naturally.
The problem? LPM 1.0 is not publicly available. miHoYo has confirmed there are no plans to release model weights, APIs, demos, or any public service for LPM. That's why we built CPMV AI — a Character Performance Model Video generator that delivers LPM-style character videos to everyone, online, powered by Veo 3.1.
Large Performance Model is a 17-billion parameter Diffusion Transformer trained on multimodal human performance data. LPM generates audio-driven character videos with full-duplex conversational behavior — speaking, listening, reacting simultaneously. However, LPM remains a research-only project with no public access, no API, and no commercial availability. The LPM team has explicitly stated the model will not be open-sourced or released as a product.
CPMV (Character Performance Model Video) is the accessible alternative to LPM. Where LPM requires a research infrastructure and audio input, CPMV works entirely from text prompts in your browser. Powered by Veo 3.1, CPMV generates expressive character performance videos with natural motion, cinematic quality, and diverse character styles — photorealistic, anime, 3D, or stylized. CPMV is live and free to start today.
Like LPM, CPMV generates characters with rich facial expressions, natural gestures, and genuine emotional depth. From subtle listening reactions to dramatic speeches — every performance feels alive and socially aware.
No training pipeline, no model downloads, no GPU required. Unlike LPM which needs a full research environment, CPMV lets you generate character performance videos in minutes from a simple text prompt.
Powered by Veo 3.1, one of the most advanced video generation models available. Create LPM-quality character videos with cinematic lighting, smooth motion, and photorealistic or stylized visuals.
LPM focuses on realistic humans. CPMV goes further — generate photorealistic people, anime characters, 3D-rendered avatars, cartoon figures, game NPCs, or any style you can describe in a prompt.
Generate characters that speak, listen, react, and emote — capturing the complete performance spectrum that LPM defined. The character behaves as a participant, not just a talking head.
No downloads, no local GPU, no Python environment. CPMV runs entirely in the cloud. Works on any device — desktop, tablet, or phone. The LPM experience, accessible from anywhere.
LPM 1.0 set the standard for character performance video generation. But since LPM isn't available to the public, CPMV AI provides the closest online experience to LPM's capabilities. Here's how they compare side by side:
| Feature | LPM 1.0 (miHoYo) | CPMV AI (Try Free) |
|---|---|---|
| Public Access | ✗ Closed research only | ✓ Open to everyone |
| Online Generation | ✗ No website, API, or demo | ✓ Generate in browser |
| Input Method | Audio-driven (requires audio) | Text-to-video (describe in words) |
| Character Performance | ✓ Full-duplex conversation | ✓ Prompt-driven expression |
| Character Styles | Realistic humans only | ✓ Any style (realistic, anime, 3D) |
| Video Quality | ✓ Research-grade (17B params) | ✓ Veo 3.1 (state-of-the-art) |
| Setup Required | Research infrastructure + GPU | None — any browser works |
| Free Tier | ✗ Not available | ✓ Free daily credits |
| Commercial Use | ✗ Academic only | ✓ Commercial-friendly |
| Open Source | ✗ Will not be released | Cloud service (no setup needed) |
LPM introduced the idea that AI characters should be performers, not puppets. CPMV brings that LPM philosophy to real-world applications — here's how creators, developers, and businesses use character performance videos generated by our LPM-alternative tool:
Generate LPM-style character videos for YouTube, TikTok, Instagram Reels, and other social platforms.
Create expressive NPC performances, cutscenes, and character trailers — inspired by LPM's game NPC vision.
Produce character-driven ad creatives and AI spokesperson videos with natural, LPM-grade expressions.
Build virtual instructors with engaging, expressive teaching performances that keep students focused.
Give conversational AI assistants a visual body — the core use case that LPM was designed to power.
Create AI-driven VTuber or digital human content for live streaming — an LPM target application.
Rapidly generate character performance scenes to plan shots and blocking before full production.
Add AI character performances to pitch decks, training videos, and corporate communications.
Write a text prompt describing the character, their appearance, emotional state, and the performance you want. Be specific about the scene — a speech, a reaction, a conversation, a greeting — CPMV understands LPM-style performance descriptions.
CPMV's video engine (powered by Veo 3.1) processes your prompt and generates a high-quality character performance video with natural motion, facial expressions, gestures, and cinematic composition.
Preview your character performance video, download it in high resolution, and use it for content creation, game development, marketing, social media, or any project that needs expressive AI characters.
While LPM 1.0 uses a 17-billion parameter Diffusion Transformer trained on human performance data, CPMV AI takes a different but equally powerful approach. CPMV is built on Veo 3.1, Google's state-of-the-art video generation model, which excels at producing high-fidelity video with natural character motion, accurate physics, and cinematic quality.
The key difference: LPM requires audio input to drive character performance (audio-to-video), while CPMV uses text prompts (text-to-video). This makes CPMV more accessible — you don't need to prepare audio files or set up a speech synthesis pipeline. Simply describe the character and their performance in natural language, and CPMV generates the video.
Both approaches share the same fundamental goal that LPM defined: generating characters that appear as expressive, socially aware performers — not just animated faces. CPMV captures the LPM philosophy of full-spectrum character performance and makes it available to everyone through a simple web interface.
CPMV AI — the online Large Performance Model alternative. Create expressive character videos from text prompts. Powered by Veo 3.1. Free to start.
Start Generating — It's Free →