FAQ
Questions about Echo and personas
How named personas work, pricing tiers, Fundamental deliverables, and privacy.
Default routes to DeepSeek V4 Flash for most drafts. For long-form content (future), it will use DeepSeek V4 Pro. You can also bring your own API key for OpenAI, Anthropic, xAI, Google, Kimi, Qwen, or DeepSeek.
Every free account starts with 50,000 tokens, enough for about 35-45 email drafts. Tokens never expire. When your pool runs out, upgrade to Pro for unlimited drafting or switch to your own API key.
Start with Perennial Echo. Create or import a persona .md, pick a channel target, paste context, and copy drafts into Mail, Slack, LinkedIn, or wherever you write. Built-in MCP drafts in the active persona inside assistants you already use.
The implementation guide walks through Claude Projects, Custom GPTs, local models, and other stacks if you prefer not to use Echo.
- Ghost-write for clients in their voice: one persona per client or executive
- Switch personas when you change inboxes or roles
- Run support and CX in a governed team persona
- Draft fluent professional English from an ESL persona you define once
- Keep channel output consistent without re-prompting every time
A persona is a markdown file that encodes how someone writes: a client, executive, brand, or team voice. Create one in Settings, start from Echo's template, or import a communication fingerprint from Perennial Fundamental.
Free includes 1 persona slot; Pro (5) and Ultra (20) unlock more per profile. Switch the active persona from the workspace without reopening Settings.
Writing your own prompt produces a described version of a voice. Perennial Fundamental extracts actual communication patterns from calibrated questions and conversation threads designed to capture unguarded language.
Self-report gaps, unconscious micro-syntax, and caricature over-saturation are why DIY prompts often feel off. The pipeline is built to reduce those failure modes.
Depends on depth level: Basic 35-50 minutes, Standard 55-75 minutes, Detailed 80-105 minutes. Progress saves every 5 questions. Resume anytime.
Email us. We will review the pipeline output and reprocess your questionnaire or issue a full refund. No explanation required.
If the delivered persona does not feel accurate, we have not done our job. Feedback improves output quality for every new customer.
A communication fingerprint is a Markdown system prompt calibrated to a specific voice. On perennial.you it is the Fundamental deliverable; in Echo you import it as a named persona.
We keep fingerprint as a secondary SEO and technical term. Day-to-day product language is persona and persona slots.
Echo Free is $0 forever (1 persona slot). Echo Pro is $3.99/mo (5 slots). Echo Ultra is $9.99/mo (20 slots).
Perennial Fundamental is a one-time $22 USD purchase for a research-backed persona file. No subscription for the Fundamental deliverable.
If a Fundamental result does not feel accurate, contact us for a refund or redo per the Terms and Privacy policies.
Install Perennial Echo first. Import the .md in Settings → Personas, set it active, choose a channel, paste context, and copy the draft out. Enable built-in MCP for the same persona in assistants.
Because you own portable markdown, you can also deploy it in Claude Projects, a Custom GPT, Grok, OpenClaw, Hermes Agent, or local models via Ollama, Open WebUI, or LM Studio.
The methodology is grounded in research across psycholinguistics, personality science, computational linguistics, and AI persona simulation.
External validation includes Park et al. (Stanford / Google DeepMind) and Toubia et al. on LLM-based digital twins. Questionnaire items map to BFI-2, Schwartz values, and LIWC. The Security & Integrity Layer follows OWASP Top 10 LLM Security standards (2025).
| Research | Authors / Year |
|---|---|
| Linguistic Inquiry and Word Count (LIWC) | Pennebaker & King, 1999; Tausczik & Pennebaker, 2010 |
| Big Five / BFI-2 | Soto & John, 2017 |
| Theory of Basic Human Values | Schwartz, 1992; PVQ-RR validated in 49 cultures |
| Self-other agreement meta-analysis | Kim et al., 2019 (n=33,000+) |
| Informant-reports superiority | Kolar, Funder & Colvin, 1996 |
| Generative Agent Simulations of 1,000 People | Park et al., 2024 (Stanford/Google DeepMind) |
| Twin-2K-500 dataset | Toubia et al., 2025 |
| TwinVoice benchmark | Li et al., 2025 |
| PersonaLLM | Hu et al., 2024 |
| Emoji and personality | Bai et al., 2019; Kennison et al., 2024 |
| Communication Accommodation Theory | Giles, 1970s-present; Burgoon et al., 1995 |
| Stylometry and authorship attribution | Stamatatos, 2009; Hollingsworth, 2012 |
| Narrative identity | McAdams; Khanal, 2025 |
| Prompt injection security | OWASP Top 10 LLM (2025); MDPI 2026 |
| PERSONAGE NLG system | Mairesse & Walker, 2010 |
| Childhood to adult personality | Longitudinal studies; Erikson's developmental theory |
Your answers are used solely to generate your persona file. We do not train on your data, sell it, or share it.
Raw questionnaire answers are automatically deleted 30 days after delivery. The file emailed to you is yours.
We use the paid xAI Grok API for processing. xAI states that API inputs and outputs are not used to train models by default.
To build an accurate persona file, we collect markers that define linguistic roots: birth year, profession, and regional history. We never ask for sensitive identifiers like home addresses, government IDs, or financial information.
You choose how much context to provide by selecting depth level. Basic requires only a continent and birth year.
Yes. We collect only the markers necessary to calibrate a persona. Every delivered file includes a Security and Integrity Layer built to OWASP 2025 LLM Security standards.
Raw questionnaire data is deleted 30 days after delivery. Echo personas and drafts stay on your machine unless you send text to an AI provider you configure.