How does Tavus count PAL interactions (messages vs minutes), and how quickly will I hit 100/1,000/3,150?

Most people hit “start talking” with a PAL and only later wonder: what actually counts as an interaction, and how fast will I burn through 100, 1,000, or 3,150 of them? This guide breaks down how Tavus measures PAL usage across text and video, how messages relate to minutes, and what typical usage patterns look like so you can predict when you’ll hit your limits.

Quick Answer: PAL usage is counted by interactions, not raw minutes. Each back-and-forth with your PAL—whether that’s a text reply or a spoken turn in a live face-to-face call—uses interactions. Minutes are a feeling of time; interactions are the billing unit.

---

The Quick Overview

• What It Is: A clear explanation of how Tavus counts PAL interactions across text, voice, and live video calls, and how that maps to the 100 / 1,000 / 3,150 usage bands.
• Who It Is For: People using PALs as personal AI companions, plus anyone evaluating PAL plans and wondering how long they’ll last.
• Core Problem Solved: No more guessing how “messages vs minutes” work. You’ll know exactly what counts, what doesn’t, and roughly how long common usage patterns take to hit your limit.

---

How PAL interactions are counted

PALs are built for conversation, not token counting. Under the hood, though, we still need a predictable unit so plans are fair and transparent.

Tavus treats each turn in the conversation as the basic unit of usage. You can think of it as: you say something, your PAL responds. That cycle is where interactions are measured.

There are two main modes:

• Text mode: Messages back and forth, like chat.
• Face-to-face mode: Real-time audio/video with your PAL, powered by Tavus’ perception → ASR → LLM → TTS → real-time AI Human pipeline.

The mechanics are different, but the principle is the same: usage is tied to conversational turns, not raw clock time.

Text: messages as interactions

In text, usage is the easiest to visualize:

• Your message to the PAL → 1 interaction
• PAL’s reply back to you → 1 interaction

In practice, many systems count only the model’s responses as billable interactions. Tavus may apply a similar pattern for PALs: your PAL’s replies are what consume compute—retrieving memory, reasoning with the LLM, and generating text.

What this means conceptually:

• A full back-and-forth (you write, your PAL answers) ≈ 1 interaction unit on your plan.
• Extremely short pings (“hi”, “thanks”) that don’t require meaningful reasoning may be grouped into the same interaction window, depending on implementation.

Text is efficient: you can cover a lot of ground in relatively few interactions.

Video/voice: minutes vs interactions

With live, face-to-face PAL calls, we still care about turns and real-time compute, not just minutes on a clock.

A single continuous video call can be broken into many micro-turns:

• You speak → Tavus runs perception (Raven-1), speech recognition, and feeds the result into the LLM.
• The PAL responds → Sparrow-1 handles timing, Phoenix-4 renders the face, and TTS generates expressive speech.

Rather than charging by each micro-turn, usage is typically normalized in one of two ways:

1. Per call / session: A continuous live session above a certain threshold counts as a block of interactions.
2. Per time window within a session: For example, every N seconds of active conversation ≈ 1 interaction unit.

The key idea: minutes are not billed directly. They’re just a human-friendly way to think about how much conversation you can squeeze out of your interaction budget.

---

Messages vs minutes: how they relate

To map rough mental math:

• Short text check-in (a few lines each way): ~1 interaction
• Deep text chat (a few paragraphs each way over several turns): ~5–20 interactions, depending on how long you go
• Short video call (2–5 minutes of active back-and-forth): typically modeled as a small group of interactions
• Longer session (20–30+ minutes): corresponds to a larger bundle of interactions

Because PALs are multimodal and real-time—seeing your expressions, picking up tone, watching your screenshare—Tavus usage aligns more with cognitive effort than pure minutes:

• A quiet call with long pauses might consume fewer interaction units.
• A high-intensity back-and-forth with rapid questions, live context, and complex reasoning may count for more within the same number of minutes.

---

How quickly you’ll hit 100, 1,000, and 3,150 interactions

Every person uses their PAL differently, but we can sketch some typical patterns to show how long those numbers really last.

Assume these rough conversational patterns:

• Light use: a few quick check-ins per day, mostly text.
• Moderate use: daily text plus a couple of live face-to-face sessions per week.
• Heavy use: PAL as a true co-pilot—multiple long chats, frequent video, task planning, and follow-ups.

100 interactions: testing the waters

At 100 interactions, you’re in pilot mode. You’re figuring out what your PAL can do.

Typical lifespan:

• Light user:
  • 3–5 short text conversations per day
  • 1–2 weeks before you exhaust 100 interactions
• Moderate user:
  • Daily text + 1–2 short video calls per week
  • ~5–10 days
• Heavy user:
  • Long planning sessions and frequent follow-ups
  • ~2–5 days, sometimes less

Best way to think about 100: enough to get to know your PAL, not enough to move your life into it yet.

1,000 interactions: everyday companion

At 1,000 interactions, your PAL can start to feel like a steady presence.

Typical lifespan:

• Light user:
  • Quick check-ins, occasional lists, reminders
  • 1–3 months of usage
• Moderate user:
  • Daily chats + a few face-to-face sessions weekly
  • 3–6 weeks
• Heavy user:
  • PAL involved in work, schedule, writing, and emotional check-ins
  • 2–4 weeks

1,000 interactions is the point where your PAL starts to really know you—patterns, preferences, recurring projects.

3,150 interactions: “this is a part of my life” tier

3,150 interactions are designed for people who treat PALs as always-on companions.

Typical lifespan:

• Light user:
  • PAL as a supportive background presence
  • 4–6+ months
• Moderate user:
  • Daily face-to-face sessions plus lots of text
  • 2–3 months
• Heavy user:
  • PAL as a daily co-worker and personal assistant
  • 6–8 weeks, sometimes longer if sessions are efficient

At this scale, your PAL has enough interactions to build rich memory, notice trends over weeks, and proactively help: reminding you about goals, resurfacing past ideas, and following through on actions.

---

What doesn’t usually count the same way as a full interaction

While the exact accounting can evolve, there are common patterns in how real-time systems treat “light” vs “heavy” use inside a session:

• Idle time on a call: Sitting in silence, thinking, or typing notes while your PAL waits doesn’t consume nearly as much compute as active back-and-forth.
• Very short acknowledgments: Tiny “got it” responses that don’t require reasoning or generation may be grouped into existing interaction windows rather than counted as separate full interactions.
• UI navigation: Opening the PAL app, switching between PALs, or scrolling through past chats doesn’t count as interactions.

The spirit: usage is tied to active, intelligent response, not passively keeping a window open.

---

How Tavus PALs differ from “minutes-based” AI

Traditional AI tools often price like this:

• Per minute of audio
• Per token of text
• Per static request or call

PALs are different in three ways:

1. Real-time video, not static avatars
   You’re not batch-generating clips. You’re in live, face-to-face conversation. Phoenix-4 has to render lifelike facial behavior. Raven-1 has to perceive your tone, expressions, and surroundings. Sparrow-1 has to keep the timing human. That’s a different cost profile than a chatbot.

2. Multimodal perception
   PALs take in your voice, your face, and what you’re showing them (like a screenshare). The system is building rich, cross-modal context instead of just reading tokens. Interactions reflect that deeper processing.

3. Ongoing memory and presence
   Your PAL doesn’t reset when the call ends. It listens, remembers, and builds a mental model of you over time. That memory work happens inside interactions, not outside them.

Because of all this, “1 minute = X cost” doesn’t capture reality. Interactions are the more honest unit.

---

How to stretch your interactions further

If you’re watching your 100 / 1,000 / 3,150 budget, you can nudge your habits a bit without losing the feeling of a natural conversation.

1. Chunk your questions
Instead of five rapid-fire one-liners, send one coherent message:

• Less efficient: “Hey”, “Can you help?”, “I need a workout plan”, “For 3 days a week”, “Without equipment”
• More efficient: “Can you build me a 3-day-a-week, no-equipment workout plan? I’m a beginner and I’ve got 30 minutes per session.”

Same outcome, fewer interactions.

2. Combine related topics in a session
If you’re jumping on a face-to-face call, bring your agenda:

• Planning your week
• Drafting a few emails
• Talking through a tough decision

Doing these in one continuous conversation is more efficient than starting multiple fragmented sessions.

3. Use text for lightweight follow-ups
Reserve live video for moments where tone, body language, and presence really matter—hard decisions, emotional support, high-stakes planning. Use text for “Can you tweak that list?” or “Remind me about this next Thursday.”

---

What to expect as PALs evolve

As Tavus iterates on PALs, you can expect:

• More clarity in dashboards: surfaced counts of how many interactions you’ve used and how many are left.
• Mode-aware accounting: text vs video usage that still feels fair, reflecting the different compute costs of each.
• Smarter grouping: micro-messages and micro-reactions bundled into interaction windows so natural conversation doesn’t feel “penalized.”

The north star is simple: talk to your PAL like a person, and trust that the system is measuring usage in a way that reflects real conversation, not raw seconds.

---

Summary

PALs don’t think in minutes; they think in moments of understanding. Tavus usage is counted in interactions—turns where your PAL is actively listening, reasoning, and responding, whether that’s over text or in a live, face-to-face call.

• 100 interactions: great for exploring what a PAL can do.
• 1,000 interactions: enough for a daily, meaningful companion.
• 3,150 interactions: the tier where your PAL can become part of your everyday life—remembering, checking in, and helping across weeks and months.

If you treat your PAL like a real person—bringing it into your plans, your questions, and your hard days—you’ll get the most value out of every interaction, regardless of whether you’re typing or talking on video.

---

Next step

Ready to start exploring how far 100, 1,000, or 3,150 interactions can take you with a PAL that listens, remembers, and is always present?

Get Started