When to consider Quark LiveAvatar
Consider Quark when you want to experiment with a larger streaming diffusion system and can evaluate its hardware requirements. Its published real-time setup uses multiple GPUs.
Pricing by mode
An open research system for long-form audio-driven avatar video. Its 14B diffusion model reports 45 FPS on a multi-GPU H800 setup.
Real-time result uses multiple H800 GPUs. Single-GPU inference also available.
| Detail | Quark LiveAvatar |
|---|---|
| Model | LiveAvatar 1.1 |
| Full avatar usage rate | Your compute |
| Full billing basis | Your compute + conversation stack |
| Video-only usage rate | Your compute |
| Video billing basis | Your GPU costs |
| Full response timing | Your pipeline · Full response depends on your stack |
| Video timing | 45 FPS · Multi-card H800, 4-step sampling |
| Output resolution | Configurable · Depends on inference settings and GPU memory |
| FPS | 45 FPS · Multi-card H800, four-step sampling |
| Custom motions | Not verified · Audio-driven diffusion; custom signature controls not established |
| Moderation control | App controlled · Your app sets filters; model and license restrictions still apply |
| Deployment | Self-hosted |
Usage rates are one component of cost. Use the monthly calculator for supported plan allowances and session assumptions.
Monthly cost examples
A monthly total needs a verified plan quote or measured compute budget. Use the provider or project sources to establish billable units, preparation, and delivery costs before estimating.
Plan the integration
Pin the repository and weights, record inference settings, and measure the full path through video encoding and delivery. Treat GPU availability and reported throughput as separate questions.
Read the latency correctly
Single-GPU support does not imply the reported multi-H800 speed. Output resolution and GPU memory depend on the inference configuration.
Single-GPU availability does not imply the reported multi-GPU real-time speed. This Quark project is separate from HeyGen LiveAvatar.
Questions to verify before launch
- Which weight and code version will the run use?
- Does your hardware match the published real-time setup?
- What is the memory requirement at the chosen resolution and concurrency?
Official sources
Quark LiveAvatar
Continue comparing
Explore all active models · Read the methodology · Explore integration and demo guides