Runway AI 2026: Gen-4, Gen-4.5, Act-Two and Aleph

TL;DR. Runway is a set of AI models for images and video. In the Pixyn catalog there are four video models: Gen-4.5 and Gen-4 Turbo (animating a frame into a short video, image-to-video), Act-Two (transfers facial expressions and motion from an "actor" video to your character) and Aleph (changes already shot video from text: background, style, objects). Runway's strong side is image-to-video and video editing. You can open Runway models in the Pixyn catalog. Below: what to take for which task, a working prompt and a comparison with alternatives.

---

What Runway is

Runway is a studio that makes generative models for images and video (the Gen-4 lineup and the Act-Two and Aleph tools). It is known first of all for video: Runway was one of the first to give directors and marketers text-to-video and image-to-video without a film crew.

The key idea: "the Runway neural network" is not one button but several different models for different tasks. If you pick the wrong one, the result will disappoint. So let us go through each.

In Pixyn the Runway models are connected through the official API. This is an explainer article; if you are already choosing between video models, we have separate comparisons, with links at the end.

Which Runway models exist and what each one does

Runway Gen-4.5 and Gen-4 Turbo (video)

This is image-to-video: you give a starting frame (a product photo, a portrait, an illustration), and Runway animates it into a short clip with the specified motion. Gen-4.5 is for quality, Gen-4 Turbo for speed and drafts. The main use: animate a still frame, make a dynamic cover for Reels/Shorts, "turn" a product.

Runway Act-Two

Transfers a performance (facial expressions, head and body movements) from an "actor" video to your character. You shoot a short video where a person speaks and moves, and Act-Two applies that performance to a generated or uploaded character. This covers the request "animate an avatar and make it talk" without classic lip sync from audio.

Runway Aleph (video-to-video)

Edits finished video from a text instruction: change the background and surroundings, recolor, add or remove an object, change the style. In essence it is "an editor by text", but for video rather than photos. Convenient when the clip is shot but you need to change the scene without a reshoot.

Model	Type	What it does	Good for
Gen-4.5	Video (image-to-video)	Animates a frame, focus on quality	Final short clips
Gen-4 Turbo	Video (image-to-video)	Animates a frame, focus on speed	Drafts and the first video
Act-Two	Video (performance)	Transfers expressions and motion to a character	A talking avatar without lip sync
Aleph	Video (video-to-video)	Changes background, style, objects from text	Editing shot video without a reshoot

The table's takeaway: take Gen-4 Turbo for drafts and Gen-4.5 for the final video, and Act-Two and Aleph to animate a character and edit a finished clip.

How to use Runway in Pixyn, step by step

1. Sign up for Pixyn: 30 seconds via email, Telegram or Google.
2. Open the model catalog and choose the Runway video model you need.
3. Upload a starting frame (for image-to-video) or a video reference (for Act-Two / Aleph) and write a prompt.
4. Get the result and regenerate with a refined description if needed.

Which tasks to use Runway for

• Marketplaces. Animate a card: a still product photo becomes a short video with light camera movement. Video in a card raises engagement.
• Reels / Shorts / clips. Dynamic covers and short AI inserts from a single frame.
• Content with a "talking" character. Act-Two: an avatar or a hero repeating the expressions and speech of a take you shot.
• Editing shot video. Aleph: change the background, season or style without a reshoot.

A prompt that works

For image-to-video, describe the motion and the camera, not the picture itself (it is already in the starting frame):

Slow cinematic push-in on the product, soft studio light, subtle rotation,
shallow depth of field, smooth natural motion, 5-second loop

Runway understands a prompt in English noticeably better: the model was trained mostly on English descriptions.

Anti-prompt (don't do this)

beautiful product video, very cool, lots of motion, dynamic!!!

Why it does not work: there are no specifics about the motion and the camera, and Runway ignores exclamations and evaluative words. The more precisely the motion is described, the more predictable the result and the fewer trial generations.

Runway or alternatives

Runway is not the only video model in the catalog, and in some scenarios competitors are stronger:

• If you are choosing between video models by quality, speed or portraits, read the Kling guide.
• If the task is not video but editing a photo with text, look at the Reve 2.0 editor: changing the background, objects and clothes on a finished shot is easier there.

For long story clips and pure text-to-video, other models are often chosen; Runway shines where there is a starting frame or a source video.

What it costs in tokens

Runway: from 25 tokens per clip, Kling: from 15, Seedance: from 20. Runway is in the expensive part of the video catalog, so drafts are cheaper to run on Seedance.

The studio shows the exact charge before you run it, so check against it rather than the article: the catalog gets updated. Pack contents are on the pricing page.

FAQ

What is the Runway neural network in simple words? It is a set of AI models from one studio for images and video. In Pixyn: animating a frame into video (Gen-4.5 / Gen-4 Turbo), transferring expressions to a character (Act-Two) and editing finished video from text (Aleph).

How is Runway Act-Two different from regular lip sync? Act-Two transfers to the character not only the lips to the sound but the whole performance: expressions and movements from the video take. It is closer to "animate and act" than to syncing lips to audio.

What does Runway Aleph do? It edits already shot video from a text instruction: changes the background and style, adds or removes objects, without a reshoot.

Is Runway suitable for marketplace cards? Yes, for short videos from a product photo (image-to-video). A long clip has to be assembled from several generations.

Which Runway model is better to start with? For the first video, Gen-4 Turbo (fast and cheap in tokens for a draft), and the final on Gen-4.5.

---

Ready to try it yourself? Open the Runway models → Sign-up takes 30 seconds via email, Telegram or Google.

Plans and balance are on the /pricing page. If you plan many similar generations, look at the model catalog: it also has the Kling, Veo and Seedance video models for comparison on your task.

For reference about the company: Runway on Wikipedia.