# Runway Act-Two: How to Drive a Character With Your Performance

> How Runway Act-Two works in November 2026: driving-video rules, image vs video characters, voice swaps, multi-character dialogue, and 5 credits a second.

_Source: https://professionalstoolkit.com/articles/runway-act-two — The Professional's Toolkit · updated November 2026_

---

Runway Act-Two turns a video of you acting into a performance by a different character. You record one person speaking and moving, add a character as an image or a video, and Runway carries the head, face, body and hand motion across to it. The tool runs in the Runway web app only, needs the Standard plan or higher, and stops at 30 seconds per generation. It costs 5 credits a second with a 3-second minimum [1], so a full 30-second take is 150 credits by our arithmetic. Pick Act-Two when you can act the line yourself. Skip it for two speakers in one shot.


**30 s** — Longest Act-Two performance per generation, one character at a time (Standard plan or higher).



- **5 credits/s** — Act-Two rate in the web app, 3-second minimum
- **1 character** — Per generation; dialogues are composited
- **24 fps** — Output, 1280x720 at 16:9


## Before you open it: plan, browser and two inputs

Act-Two lives in the **Apps section** of the web app, and it is not in the iPhone or Android apps, so plan to work from a desktop browser. The Free plan cannot run it. Standard, at $15 a month or $12 a month billed yearly, is the entry point, and every plan above it includes the tool. Runway launched Act-Two in July 2025 as the successor to Act-One, which had arrived in October 2024 inside Gen-3 Alpha; Gen-3 itself was retired in mid-2026, so older tutorials that start from the Gen-3 screen no longer match what you will see.

Every generation needs exactly two inputs. The first is the **driving performance**: a clip of a real person delivering the line, which supplies motion, expression and audio. The second is the **character**, either a still image or a short video. The character can be your own artwork, a photo you hold the rights to, or something generated inside Runway; the help center notes that a Gen-4 Video output can be reused as the character or the performance clip. Runway also lists an Add Performance app that wraps the same job, taking a 1 to 30 second performance and an optional voice from the voice library.

## Recording a driving take that transfers cleanly

Runway's rules for the performance clip are short and strict. Film a **single subject** whose face stays visible the whole time. Frame no wider than waist up. Keep it to one continuous shot with no cuts. Some expressions do not carry over at all; Runway names sticking out a tongue as one example. If a scene needs a cut, shoot it as separate takes and generate each one on its own.

The camera itself matters less than the framing. Runway pitches Act-Two to game teams as a way to record a driving performance with any camera and move it onto a reference character, with no rigging, hand animation or motion-capture shoot. That claim is Runway's, but the input rules support it: nothing in the help article asks for markers, depth sensors or a studio. What it does ask for is a readable face, because the face drives the lip shapes and the expressions you will judge first.

Length deserves a plan before you press record. A take under 3 seconds is still billed as 3 seconds, or 15 credits, so very short pickups waste money. Anything over 30 seconds will not fit one generation. Rehearse to a clean read, then trim the clip to the delivery before uploading.


> 💡 **Edit the scene before the performance:** Aleph is not a lip-sync model, so environment edits made after Act-Two can force a second Act-Two pass. Do Aleph work first.


## Character image or character video: what changes

The choice of character input decides who controls the camera. With a **character image**, Act-Two invents environmental and camera motion around the figure and lets you turn on gesture control, so hand and body movement from your take reach the character. With a **character video**, the original camera and scene motion stay as they were, and gesture control is unavailable. If your performance runs longer than the character video, Runway loops the video in a reversed boomerang to fill the time, which can look odd on a moving background. Match the lengths when you can.

A facial expressiveness slider runs from 1 to 5, with 3 as the default. Lower settings tone the acting down but can hold the character's likeness more steadily; higher settings push emotion and risk visual artifacts. Start at the default, judge one take, then move a single step at a time. For a recurring character, Runway's guide on longer films recommends character plates: neutral images of the same figure from several angles, outfits and framings, so each shot starts from a consistent reference. Output follows the aspect ratio you choose, always at 24 fps:

| Aspect ratio | Act-Two output |
|---|---|
| 16:9 | 1280x720 |
| 9:16 | 720x1280 |
| 1:1 | 960x960 |
| 4:3 | 1104x832 |
| 3:4 | 832x1104 |
| 21:9 | 1584x672 |

These are 720p-class frames. Pick the ratio before you shoot. A vertical 9:16 delivery wants a performer framed for a tall crop, while 21:9 leaves room on both sides. If you need a larger master, paid plans include 4K upscaling as a separate step. Enterprise admins get one extra control here: they can turn off face detection on Act-Two character images.



> ⚠ **Short character videos loop:** If the performance runs longer than the character video, Runway plays the character video back and forth in a boomerang loop to fill the time.

## Changing the voice and staging a dialogue

The output carries the audio of your take. To swap it, choose Use and then Change Voice on the finished clip. Expect one quirk: the new voice may keep the accent of the original speaker. A custom voice of your own is available from Pro, which allows one (Max allows three, Team five), and each costs 50 credits to create. Training one requires the speaker to read a consent script in the web app, with no exceptions, public figures included, and cloning anyone else needs their explicit permission.

**Act-Two handles one character per generation.** For a conversation, Runway's own method is to generate each character separately and then composite them in a video editor [2]. Keep the whole exchange under 30 seconds, since every speaker's take must fit one generation. When the scene needs a longer base plate, the same guide suggests extending a Gen-4 Video clip by using the last frame of a 10-second clip as the first frame of the next.

One ordering trap sits in that workflow. Aleph is not a lip-sync model, so if you re-light or restyle the environment with Aleph after Act-Two, mouth movement can drift and you may need to run Act-Two again. Do the scene edits first and the performance last; our [Aleph editing guide](/articles/runway-aleph-video-editing) covers that tool's inputs and minimums.


> ⚠ **Rejected takes still cost credits:** Runway returns Act-Two credits only for generation errors; a take with a stiff mouth or warped hands is charged in full.


## What each take costs on each plan

Act-Two shares its rate with Gen-4 Turbo: 25 credits per 5 seconds. By our arithmetic, a month of plan credits covers about 125 seconds of Act-Two on Standard, 450 on Pro and 1,900 on Max, if you spend them on nothing else. That is the ceiling, not the yield. A take with a stiff mouth or a warped hand still costs its credits, because Runway returns credits only for generation errors, so the real price of a usable performance depends on how many retakes you need. Budget for several.

If you outgrow Standard, compare the price of a credit rather than the plan price. By our arithmetic, a Standard credit works out to 2.4¢ on monthly billing, a Pro credit to 1.56¢, and a top-up credit to 1.0¢, since paid plans can buy 1,000 credits for $10. On monthly billing, topping up Standard gets you extra seconds more cheaply than upgrading. On yearly billing the gap nearly disappears. Pro earns its higher price through other things this workflow uses, such as the custom voice and 15 parallel generations.

Standard and Pro credits reset at each billing date, while top-ups never expire, which suits occasional dialogue work. Refunds are discretionary and limited to recent, nearly unused purchases; the full billing rules sit in our [Runway pricing guide](/articles/runway-ai-pricing). Developers can call Act-Two through Runway Dev too, but those API credits are a separate pool from your app plan.


> 💡 **Top up instead of upgrading early:** On monthly billing, a top-up credit (1,000 for $10, never expires) costs less than a Pro plan credit, so a slightly busier month on Standard can be covered with top-ups.


## When another tool fits the job better

Searches for Runway lip sync now land on Act-Two for a reason. The standalone Lip Sync tool was retired in May 2026, with Act-Two named as its replacement. If you have no performance to record, the **Add Dialogue** app turns a still image into a speaking character with automatic lip sync for 320 credits per video [3], though it uses a default voice for now.

Runway Characters is a different product. Built on Runway's GWM-1 world model, it generates real-time conversational avatars from a single image instead of replaying a recorded take. In the web app you talk with preset avatars for up to 2 minutes per conversation at 2 credits per 6 seconds, and even Free users can try them; custom avatars, knowledge bases and longer sessions sit on Runway Dev. Choose it for interactive demos, not for scripted scenes.

Inside the same Runway plan, **Kling 3.0 Motion Control** animates a still character image from a performance video at 1080p, priced at 85 credits per 5 seconds, in 16:9, 9:16 or 1:1. It is worth a test when you need a larger frame than Act-Two's 720p-class output. For a talking presenter rather than an acted character, HeyGen is built around avatars, with its Creator plan at $29 a month or $24 a month billed yearly; our [HeyGen review](/articles/heygen-review) covers it.

Recording faces brings obligations. You need the rights to any likeness you bring in, Runway's consumer Terms let it train on inputs and outputs (no-training is an Enterprise term), and its privacy terms say biometric data such as face scans is kept no longer than needed, or up to three years after your last interaction.


## FAQ

**Can I use Act-Two on the Runway iPhone or Android app?**

No. Act-Two sits in the Apps section of the Runway web app, and Apps are not in the iOS app; work from a desktop browser.

**Does Act-Two need a motion-capture suit or special camera?**

No. Runway says a driving performance can be recorded with any camera, with no rigging or mo-cap shoot; the face must stay visible, framed waist up at most, in one continuous shot.

**Can Act-Two make a character stick out its tongue?**

Not reliably. Runway lists some expressions, including sticking out a tongue, as unsupported in the driving performance.

**What happened to Runway Act-One and Lip Sync?**

Act-One launched in October 2024 inside Gen-3 Alpha, which Runway retired in mid-2026; the standalone Lip Sync tool was retired in May 2026 with Act-Two named as its replacement.


## The bottom line

Act-Two is the most direct route in Runway from an actor's take to an animated character: one person, one continuous shot, up to 30 seconds, for 5 credits a second on Standard or higher. It rewards preparation. Shoot clean waist-up takes, choose an image when you want gestures and a video when you want to keep the camera, finish environment edits before the performance pass, and budget for retakes rather than first tries. For multi-speaker scenes, plan the compositing from the start; for no actor at all, Add Dialogue or Characters suit the job better.



## Sources

[1] Runway Help Center: Performance Capture with Act Two — https://help.runwayml.com/hc/en-us/articles/42311337895827-Performance-Capture-with-Act-Two (2026-11)
[2] Runway Help Center: Creating Multi Character Dialogues with Act Two — https://help.runwayml.com/hc/en-us/articles/41748090660499-Creating-Multi-Character-Dialogues-with-Act-Two (2026-11)
[3] Runway app: add dialogue — https://runway.com/apps/add-dialogue (2026-11)
[4] Runway pricing (US) — https://runway.com/pricing
[5] Runway Help Center: how credits work — https://help.runwayml.com/hc/en-us/articles/15124877443219-How-do-credits-work
[6] Runway Terms of Use — https://runway.com/terms-of-use
[7] Runway changelog — https://runway.com/changelog