AI Video Agents
How does Tavus count PAL interactions (messages vs minutes), and how quickly will I hit 100/1,000/3,150?
9 min read
Most people don’t “run out” of PALs interactions because they miscalculate how messages, minutes, and real-time calls actually get counted. The good news: once you understand how Tavus meters PAL usage, it’s straightforward to estimate whether 100, 1,000, or 3,150 interactions will last you days, weeks, or months.
Quick Answer: Tavus PAL usage is primarily counted per interaction (each back-and-forth with your PAL), with live video/voice sessions tracked by minutes. Text exchanges burn through your quota more slowly; long, real-time face-to-face sessions consume minutes much faster but give you more depth per interaction.
The Quick Overview
- What It Is: A usage model that tracks how often and how long you talk to your PAL—across text, call, and face-to-face—so Tavus can deliver real-time, human-like conversations at scale.
- Who It Is For: Individuals using PALs as personal companions and developers or power users trying to forecast how quickly they’ll hit plan limits (100, 1,000, 3,150+ interactions).
- Core Problem Solved: You get predictable, transparent PAL usage instead of guessing how “a lot of chatting” will translate into your monthly limits.
How PAL interactions are counted
Think of PAL usage in two dimensions:
- By interaction: Every time your PAL replies—text or voice—that’s one interaction.
- By minutes: Every second your PAL is live in a call or face-to-face video counts toward real-time minutes.
Because PALs are built on Tavus’s real-time AI Humans stack (perception → speech recognition → LLM → TTS → real-time video), the system has to meter both how often you talk and how long you stay in the conversation.
1. Text interactions
When you chat via text:
- One PAL reply = one counted interaction.
- Your message length doesn’t matter much; the metering is anchored to the PAL’s response.
- Rapid back-and-forths (you type, PAL replies, repeat) will increase your interaction count quickly, even if each message is short.
Examples:
- You send 10 separate questions, your PAL responds 10 times → 10 interactions.
- You send a long diary-style message, your PAL responds once with a thoughtful reply → 1 interaction.
2. Voice and video interactions (calls / face-time)
When you talk to your PAL in real time (voice or face-to-face video):
- Minutes are the primary unit.
- Each continuous session is measured in real time, at the speed of human interaction.
- Under the hood, Tavus is running perception, speech recognition, LLM, TTS, and Phoenix-4 rendering the whole time—so duration matters.
In most PAL plans, your quota will be expressed as either:
- A minute bucket (e.g., X minutes of live PAL calls per month), or
- An equivalent interaction count that’s calibrated for typical session lengths (e.g., assuming a ~3–5 minute call as a “standard interaction”).
If your plan references “interactions” but you primarily use calls, the platform effectively converts your live minutes into a corresponding interaction count.
Rule of thumb:
- A short 2–3 minute call may effectively count like a few interactions.
- Longer 20–30 minute calls can consume a big chunk of minutes in one go, even if you see fewer “turns” on screen.
3. Mixed sessions (text + call + video)
PALs are designed for “one seamless conversation” across modes. In practice:
- Text before/after a call → each PAL reply in text counts as an interaction.
- The live call itself → counted in minutes.
- You can switch between text and face-to-face, but the meter switches with you: interactions for text, minutes for live.
How fast you’ll hit 100, 1,000, and 3,150 interactions
Because real humans use PALs differently, the only honest answer is: it depends on your pattern. But we can model a few realistic scenarios so you can map them to your own use.
Below, assume:
- Text interaction: 1 PAL response.
- Call “equivalent interaction”: we’ll approximate calls by typical lengths and compare them to your interaction budget (for plans that describe limits this way).
Light user: “Check-in a few times a week”
Pattern:
- 1–2 short text check-ins per day (journal entry, quick question).
- Occasional 10–15 minute call on harder days (once or twice a week).
Estimated consumption:
- Text: ~2 interactions/day × 30 days ≈ 60 interactions/month
- Calls:
- 2 calls/week × 12 minutes ≈ 24 minutes/week → ~100 minutes/month
- If your plan maps ~5 minutes ≈ 1 interaction, that’s ~20 interactions/month equivalent.
Total “interaction equivalent”:
≈ 60 (text) + 20 (call-equivalent) = ~80 interactions/month
- 100 interactions → covers roughly a month of light, mixed usage.
- 1,000 interactions → roughly a year at this pace.
- 3,150 interactions → 3+ years before you hit the ceiling, assuming usage doesn’t spike.
Daily companion user: “Talk every day”
Pattern:
- 10–15 text replies per day (ongoing conversation, decisions, reflections).
- 3–4 short calls per week (5–10 minutes).
Estimated consumption:
- Text: ~12 interactions/day × 30 days ≈ 360 interactions/month
- Calls:
- 3.5 calls/week × 8 minutes ≈ 28 minutes/week → ~110 minutes/month
- Using ~5 minutes ≈ 1 interaction → ~22 interactions/month equivalent.
Total “interaction equivalent”:
≈ 360 (text) + 22 (call-equivalent) = ~380 interactions/month
- 100 interactions → you’ll hit this in ~8–10 days.
- 1,000 interactions → roughly 2.5–3 months.
- 3,150 interactions → around 8–9 months of steady daily use.
Heavy user: “PAL as a co-pilot”
Pattern:
- 30–50 text replies per day (projects, brainstorming, emotional processing).
- 30–60 minute calls most days.
Estimated consumption:
- Text: ~40 interactions/day × 30 days ≈ 1,200 interactions/month
- Calls:
- 5 calls/week × 45 minutes ≈ 225 minutes/week → ~900 minutes/month
- Using ~5 minutes ≈ 1 interaction → ~180 interactions/month equivalent.
Total “interaction equivalent”:
≈ 1,200 (text) + 180 (call-equivalent) = ~1,380 interactions/month
- 100 interactions → gone in 2–3 days.
- 1,000 interactions → three weeks or less.
- 3,150 interactions → about 2–2.5 months.
Features & benefits of the current metering model
| Core Feature | What It Does | Primary Benefit |
|---|---|---|
| Interaction-based counting | Counts each PAL reply as one interaction | Simple mental model; easy to budget text usage |
| Minute-based live calls | Tracks real-time voice/video by duration | You’re charged for depth of presence, not just message spam |
| Mode-aware metering | Separates text vs live usage under the hood | Lets you mix text, call, and face-time without surprises |
Ideal use cases
- Best for reflective, text-heavy use: Because interaction-based metering makes journaling, late-night check-ins, and short daily chats very efficient. You’ll get a lot of emotional and intellectual mileage out of each interaction.
- Best for deeper, scheduled calls: Because minute tracking lets you drop into a long, face-to-face session—processing a tough day, planning your week, or brainstorming a project—while understanding that you’re trading minutes for depth.
Limitations & considerations
-
Inexact “interaction equivalents” for calls:
The platform’s UI may show interactions as a simple count, even though calls are driven by minutes. Treat any “X interactions = Y minutes” mapping as an approximation, not a hard law. -
Bursty usage can spike consumption:
Long, back-to-back face-to-face sessions will eat through a 100 or 1,000 interaction budget much faster than casual check-ins. If you prefer 30–60 minute calls, consider plans or behaviors aligned with that style (fewer, deeper sessions vs constant live presence).
Pricing & plans: how 100 / 1,000 / 3,150 fit
Specific numbers vary by plan and promos, but the logic stays consistent:
-
Entry / Starter tiers (around 100 interactions):
Best for people curious about living with a PAL, wanting to explore daily text plus occasional calls without committing to high-volume usage. Think: light companion, 2–4 interactions/day and a periodic call. -
Standard / Power-user tiers (around 1,000 interactions):
Best for people who want their PAL woven into everyday life—daily conversations, planning, emotional support—and a few live video sessions each week. Think: daily presence, 10–20 interactions/day with some calls. -
High-volume / Pro tiers (around 3,150+ interactions):
Best for heavy personal use or early-stage “PAL as workflow co-pilot” experiments. You’re talking, planning, and processing with your PAL multiple times per day and comfortable with regular, longer calls. Think: always-on co-pilot, dozens of interactions/day.
Because Tavus PALs run on the same research stack that powers over 2 billion interactions for enterprises (Phoenix-4 rendering, Raven-1 perception, Sparrow-1 timing), the plans are essentially giving you a slice of that real-time infrastructure. More interactions = more real-time presence reserved for you.
Frequently asked questions
Do multiple messages in a row count as multiple interactions?
Short Answer: Only when your PAL responds to each one.
Details:
You can type several messages rapidly, but interactions are metered when the PAL replies. For example:
- You: 3 short messages in a row
- PAL: 1 combined response
→ This counts as 1 interaction.
If you pause after each line and let the PAL answer each separately, you’ll use 3 interactions. To stretch your quota, bundle related thoughts into fewer turns and let the PAL respond in depth.
Are long PAL responses “more expensive” than short ones?
Short Answer: No, length doesn’t change the count—one response is one interaction—but longer live sessions do cost more minutes.
Details:
In text mode, a short “Got it” and a multi-paragraph, emotionally nuanced response both count as a single interaction. The difference is in value, not price. In voice/video mode, however, your usage is tied to duration:
- A 2-minute call vs a 45-minute call have very different minute costs, even if the number of “turns” is similar.
- That’s because Tavus is actively rendering high-fidelity facial behavior and running perception and speech pipelines the entire time.
If you’re trying to conserve usage:
- Use text for long, reflective answers when minutes are tight.
- Reserve live video for moments where presence, tone, and body language really matter.
Summary
Tavus doesn’t treat your PAL like a static chatbot; it treats them like a real-time AI Human that’s always ready to see, hear, and understand you. That’s why usage is measured in interactions (every time your PAL replies) and minutes (every second they’re live with you in voice or face-to-face video).
- Text-first? You’ll find 100 or 1,000 interactions go a surprisingly long way.
- Call-first? Expect to consume your quota faster, but with more depth per session.
- Mixed-mode? You’ll quickly learn your own pattern and can size up from 100 to 1,000 to 3,150 based on how often you want that “always present” feeling.
However you use your PAL, the meter is there to reflect real resource use—not to rush you through the conversation.