Runway Characters turns a custom character into a real-time conversational video agent, so the useful pricing question is not only “what does Runway cost?” It is “what does one avatar session cost when users stay for 1, 5, 15, or 30 minutes?”
This guide treats Runway Characters as a session-cost planning problem for product demos, support avatars, onboarding flows, and creator-facing assistants. It uses the official Runway API pricing model as the source of truth, then compares the billing shape with Tavus, D-ID, and lower-cost voice-only fallback patterns.
Official Runway Characters Pricing Inputs
Runway’s API pricing page lists API credits at $0.01 per credit. For gwm1_avatars, the listed cost is 2 credits upfront, then 2 credits per 6 seconds. The official Characters documentation describes custom conversational characters powered by GWM-1, including embedded experiences and integrations where teams control the surrounding agent stack.
Runway Characters Cost Formula
Use this calculator formula for a single real-time avatar session:
billable_6_second_blocks = ceiling(session_seconds / 6)
credits = 2 upfront credits + (billable_6_second_blocks * 2 credits)
estimated_usd = credits * $0.01The upfront charge makes very short calls relatively expensive per minute, while longer calls trend toward about $0.20 per minute after the session starts. Actual monthly spend depends on abandoned sessions, retries, concurrent usage, and whether the surrounding voice, LLM, knowledge-base, and moderation layers are billed separately.
Session Cost Scenarios
| Scenario | Duration | Runway formula | Estimated credits | Estimated API cost |
|---|---|---|---|---|
| Quick product greeting | 1 minute | 2 + (10 blocks * 2) | 22 | $0.22 |
| Short support answer | 5 minutes | 2 + (50 blocks * 2) | 102 | $1.02 |
| Guided onboarding | 15 minutes | 2 + (150 blocks * 2) | 302 | $3.02 |
| Training or sales walkthrough | 30 minutes | 2 + (300 blocks * 2) | 602 | $6.02 |
Runway vs Tavus vs D-ID Billing Shape
The main comparison is not a universal winner. It is the billing unit and workflow fit. Runway exposes a direct credit formula for the avatar video layer. Tavus packages conversational-video minutes, concurrency, replica training, and API access into plan tiers. D-ID’s API pricing packages credits, streaming minutes, embedded agents, licensing, watermarking, and avatar limits.
| Platform | Useful billing lens | What to watch before scaling | Best-fit pilot |
|---|---|---|---|
| Runway Characters | Credits per real-time avatar session: 2 upfront + 2 per 6 seconds | Session length, abandoned calls, surrounding STT/LLM/TTS/tool costs, and concurrency caps | Custom avatar experiences where teams want control around the agent stack |
| Tavus CVI | Plan plus included conversational-video minutes, API access, replica training, and concurrency | Minimum monthly plan, overage behavior, concurrent streams, recordings, and custom replica needs | End-to-end face-to-face conversation products that need packaged video-agent infrastructure |
| D-ID API | Plan credits with streaming/video minutes, embedded agents, avatar limits, watermarking, and license tier | Credit capacity, commercial license, voice clone, watermarking, storage, and agent embedding limits | Embedded talking avatar or agent flows where bundled plan limits fit the use case |
| Voice-only fallback | Usually per-minute voice, transcription, model, and telephony cost | Lower visual engagement, but simpler moderation, faster fallback, and lower media cost | Support triage, qualification, or high-volume flows where video is optional |
Hidden Cost Drivers
- Abandoned calls: short sessions still include the upfront Runway credit cost.
- Six-second rounding: plan around blocks, not exact milliseconds.
- Custom voice and persona setup: the video avatar may be only one layer of the total stack.
- Knowledge base and tool calls: retrieval, LLM calls, tool execution, and logging can exceed the avatar cost in complex workflows.
- Concurrency: a cheap single call can become expensive when 20 users start sessions at once.
- Consent and brand review: custom characters need approval, QA, and moderation before public launch.
- Fallback handling: failed avatar sessions should fall back to voice, chat, ticketing, or a human review queue.
Pilot Template for a Branded Avatar Support Flow
- Choose one narrow persona, such as product demo host, onboarding guide, or post-purchase support assistant.
- Cap the pilot at 10 to 30 sessions across 1, 5, and 15 minute scenarios.
- Track session duration, drop-off, tool-call count, fallback rate, user satisfaction, and review time.
- Set a monthly credit budget before adding public traffic or concurrency.
- Compare completion rate against a voice-only or chat-only baseline before paying for larger video-agent volume.
Where This Fits in a Makefun Workflow
Runway Characters is relevant for teams designing avatar onboarding, product demos, and creator-support experiences. If your workflow is broader AI video production, start with the AI video API guide. For agentic video planning, compare the workflow against agentic video generator patterns. If you already use Runway for generation, the Runway Agent AI video generator guide provides nearby context. For D-ID-specific video agent coverage, see D-ID agentic videos.
FAQ
How much does a 15-minute Runway Characters session cost?
Using the official API formula of 2 upfront credits plus 2 credits per 6 seconds, a 15-minute session is about 302 credits, or $3.02 at $0.01 per credit.
Is Runway Characters cheaper than Tavus or D-ID?
Not universally. Runway has a clear per-session avatar-credit formula, while Tavus and D-ID package conversational-video minutes, replicas, credits, license terms, concurrency, and other limits. Compare the billing unit against your actual pilot design.
What should I budget before a public avatar launch?
Budget the avatar video session, voice stack, LLM/tool calls, knowledge-base updates, logging, moderation, QA, consent review, fallback support, and concurrency headroom. A small pilot is the safest way to find the real blended cost.
