SCREENHELPER
that works your screen for you
AI ASSISTANT
Meet ScreenHelper: a hyper-real Digital Human with a transparent background that you can talk or chat with in 70+ languages. It listens, thinks, and operates your mouse & keyboard to finish tasks on your screen. Self-host for 100% data privacy.
HOW IT WORK
PRIVACY FIRST
Self-hosting means prompts, screenshots, and outputs can stay entirely in your network: your servers, your logs, your rules. This is a top-level selling point for regulated teams.
SPEED
Response and action speed depends on your chosen provider, but modern models like DeepSeek are optimized for high throughput and low cost. Faster than a human could click through.
MULTILINGUAL
Understands 70+ languages; voice in/out optional. Natural facial expressions, gestures, and emotion presets.
See it do a real work.
DEMO VIDEO
UNLOCK EVERYDAY PRODUCTIVITY
INBOX TRIAGE
-
prioritize
-
auto-replies
-
scheduling
ADMIN & BACK OFFICE
-
HR portals
-
ERP clicks
-
invoices
RESEARCH & REPORTING
-
Web browse via your LLM
-
Summarization
SUPPORT OPS
-
Ticket sorting
-
Knowledge lookups
POWER USERS
-
Multi-app workflows
-
Personal productivity
INBOX TRIAGE
-
Prioritize
-
Auto-replies
-
Scheduling
ADMIN & BACK OFFICE
-
HR portals
-
ERP clicks
-
Invoices
RESEARCH & REPORTING
-
Web browse via your LLM
-
Summarization
SUPPORT OPS
-
Ticket sorting
-
Knowledge lookups
POWER USERS
-
Multi-app workflows
-
Personal productivity
From triaging emails to automating admin tasks, ScreenHelper helps you work smarter across apps and workflows—powered by your chosen LLM.
02
Understanding & planning → your LLM (DeepSeek, Gemini 2/+, GPT, Claude, Grok…).
We recommend DeepSeek for cost & data-control options. Official token pricing for deepseek-chat currently: \$0.27/M input and \$1.10/M output (cache hits \$0.07/M).
Note: pricing changes announced for September 5, 2025 UTC. (api-docs.deepseek.com)
03
Screen control → ScreenHelper moves the mouse & types to complete steps you can watch.
Optional personality → choose from 50+ personas (tone, style, gestures).
Optional voice out → local robotic TTS (free) or connect ElevenLabs (your account).
Everything can also be Chat-only (no STT/TTS).
FROM VOICE TO ACTION
01
You speak or chat → local speech-to-text (or text-only).
Default local engine: Whisper large-v3; disk footprint ~1.5–1.6 GB. (huggingface.co)
HOW MUCH DOES IT COST TO RUN SCREENHELPER (LLM USAGE)?
Using DeepSeek at current official rates (deepseek-chat: \$0.27/M input, \$1.10/M output), a typical interactive workload of ~200 input tokens + 150 output tokens per second, plus 1 screenshot per second, comes to roughly:
-
If each screenshot inlines ≈ 500 tokens → ~\$1.27/hour (≈ €1.09/hour).
-
If ≈ 1,000 tokens → ~\$1.76/hour (≈ €1.51/hour).
-
If ≈ 2,000 tokens → ~\$2.73/hour (≈ €2.34/hour).
Assumptions: Fireworks’ “document inlining” converts images to tokenized input; you’re billed per token by the provider. Actual cost depends on model, image complexity, and provider settings (prompt caching can lower input costs). Official DeepSeek pricing & the “document inlining” approach are documented here. (api-docs.deepseek.com, Fireworks AI)
Notes: DeepSeek has announced pricing changes effective Sep 5, 2025 (UTC). Euro conversions above use the ECB reference on Aug 26, 2025 (1 EUR ≈ 1.1656 USD). (api-docs.deepseek.com, European Central Bank)
PRIVACY & SELF HOSTING
featured block, very prominent
Deploy your own DeepSeek on servers you control; no third-party data flows.
Zero telemetry mode; configurable local log retention; IP allow-listing; SSO.
Works with your company’s DLP, SIEM, and VPC rules.
Bring-your-own vendor keys if you prefer managed hosting. (DeepSeek V3.1 and platform access are available globally)
TRY BETA VERSION
Email us at request@avataris.ai to receive a download link. We may release a web version here in the future, so stay tuned!
For updates on new products and demo releases, follow us on LinkedIn and subscribe to our newsletter.