🎬 AI Video Generation with Runway Gen-4, Gen-3, and Aleph models

About Runway

Runway's video models share one page across three generation types. Gen-4 turns prompts or stills into five- or eight-second clips, Gen-3 Alpha is the budget prompt-only tier, and Aleph transforms a clip you already have. Each type keeps its own prompt field, durations and inputs.

Open Runway when the source material already exists. Aleph takes a publicly reachable video URL and a transformation prompt, then restyles the footage: watercolour, illustration, a different grade, a different mood. It works from the clip you give it rather than from a blank prompt, so a finished edit can be re-rendered in a new look without rebuilding it shot by shot. That makes it the only generation type on this page that starts from a moving source.

Gen-4 is where most first runs should go. Text to video accepts a description of up to 1,800 characters and returns five or eight seconds at 720p, with 1080p available on five-second clips. Image to video takes the same two lengths plus a JPG or PNG up to 10MB and animates it. Gen-4 exposes four aspect ratios, 16:9, 9:16, 1:1 and 4:3, which covers landscape edits, vertical feeds and square placements from a single prompt.

The tier below is Gen-3 Alpha, labelled as the budget option: text to video only, five seconds, 720p, with 16:9, 9:16 and 1:1. Use it to check whether a prompt reads before you move up to Gen-4, then keep the settings and pay for the better render. Cost follows that choice rather than sitting at one flat number, because generation type and clip length both move the credit total. The figure is displayed before the request goes out, and runs that end without a video are refunded.

Decide the clip length before anything else, because it decides the resolution you can offer. On Gen-4, 1080p is available only on five-second clips, while the eight-second option returns 720p. A client who needs a full-HD deliverable therefore needs a five-second shot, not a longer one scaled down afterwards. The same ordering applies when the starting point is a still: Gen-3 is text only, so when a photograph is the starting point it has to go through Gen-4 image to video, with a JPG or PNG under 10MB.

Aleph is the type to reach for once the edit is already approved. It takes a publicly reachable video URL rather than an upload, so the source has to be hosted somewhere the model can fetch it, and it returns that clip in a new treatment without making it longer. The prompt allows up to 1,800 characters, which is enough to describe a grade, a medium and a mood in separate sentences. When the problem is duration rather than look, the Runway Extend workflow on the video-to-video page covers continuation.

How Runway works

1. 1

   Choose a generation type

   Aleph restyles a video URL, Gen-4 text to video starts from a description, Gen-4 image to video animates a still, and Gen-3 is the budget text-to-video tier. The form changes with the choice.

2. 2

   Supply the input

   Gen-4 text and Gen-3 need a prompt under 1,800 characters. Image to video needs a JPG or PNG URL up to 10MB. Aleph needs both a transformation prompt and a publicly accessible video URL.

3. 3

   Set length, ratio and resolution

   Gen-4 offers five or eight seconds, 720p or 1080p, and four ratios; 1080p applies to five-second clips. Gen-3 is fixed at five seconds, 720p and three ratios. Generate, then download the MP4.

What Runway is best for

• Restyling finished footage

  Point Aleph at an existing MP4 and ask for a new treatment. The cut, timing and camera work survive while the surface changes.

• Prompt tests before a paid render

  Draft on Gen-3, confirm the idea reads, then rebuild the same description on Gen-4 with more detail and a longer duration.

• Still to motion

  A JPG or PNG becomes a five- or eight-second beat with described movement, enough for a social cutdown or a hero background.

• Square and vertical placements

  Gen-4's four ratios let one idea render for a feed, a story and a square placement instead of cropping a landscape master.

Runway at a glance

Models

Runway Gen-4, Gen-3 Alpha and Aleph

Generation types

Text to video, image to video, video to video

Clip length

5 or 8 seconds on Gen-4; 5 seconds on Gen-3

Resolution

720p or 1080p; Gen-3 is 720p, and 1080p on Gen-4 applies to 5-second clips

Aspect ratio

16:9, 9:16, 1:1 and 4:3 on Gen-4; 16:9, 9:16 and 1:1 on Gen-3

Output

MP4 with H.264 encoding

Runway FAQ

Which Runway model should I use?

Aleph when you already have footage and want a new look, Gen-4 text to video when you are starting from a description, Gen-4 image to video when a still should move, and Gen-3 when you want the budget text-to-video pass before spending more.

What does Aleph need from me?

A publicly accessible video URL, with MP4 recommended, and a transformation prompt of up to 1,800 characters describing the style, effect or atmosphere you want. Aleph does not generate from a blank prompt; it needs the source clip.

How long are Runway clips?

Gen-4 returns five or eight seconds and Gen-3 returns five. There is no longer single-pass duration on this page, so a longer sequence means generating the pieces and joining them in an editor.

What resolutions and ratios can I get?

720p on every generation type, plus 1080p on five-second Gen-4 clips. Gen-4 offers 16:9, 9:16, 1:1 and 4:3, while Gen-3 offers 16:9, 9:16 and 1:1. Aleph keeps the framing of the clip you supply.

What does a run cost, and what happens if it fails?

The total depends on the generation type and, on Gen-4, the clip length. The credit figure appears in the form before you generate. A run that finishes without a video is refunded, and monthly plans bundle credits, with Starter at $4.90 a month for 300.

Can I have 1080p and eight seconds at the same time?

No. On Gen-4 the 1080p option applies to five-second clips, and the eight-second option returns 720p. Pick whichever the deliverable actually needs: a short full-HD master, or a longer clip at 720p. Gen-3 is fixed at five seconds and 720p, so it cannot help with either, and Aleph changes the treatment of a clip rather than its length. For footage that has to run past eight seconds, the continuation workflows live on the video-to-video page.