Runway ML Tutorial: How to Create AI Videos From Scratch to Pro
When Runway ML released the first version of Gen-2 in 2023, AI-generated videos felt more like prototypes than finished products. By 2026, with Gen-4 and the experimental Aleph, the story has changed. Film studios use Runway in pre-production, agencies create entire commercials without real shooting, and social media creators produce viral clips from a single image.
This tutorial walks through how to use the platform from scratch, the tools that actually move the needle (Gen-4, Image-to-Video, Lip Sync, Motion Brush, Aleph), the available plans, and the main use cases in real production.
What Is Runway ML
Runway ML is an artificial intelligence platform focused on audiovisual creation, founded in 2018 by Cris Valenzuela, Anastasis Germanidis, and Alejandro Matamala (ex-NYU ITP). By 2026, the company hit a valuation above $3 billion, with investment from Google, NVIDIA, and Salesforce Ventures, and signed strategic partnerships with Lionsgate (a landmark licensing deal to train models on the studio’s film library).
Unlike tools focused on “making a pretty video from a prompt,” Runway positions itself as AI Magic Tools for professionals: video editors, directors, agencies, and studios. The platform combines AI generation with classic editing functions (rotoscoping, green screen, color grading, motion tracking) in a single web interface.
How to Get Started
Entry is simple and runs entirely in the browser:
- Go to runwayml.com
- Create an account with Google, Apple, or email
- You get the Free plan with 125 one-time credits (limited use)
- On the dashboard, choose the generation type: Text to Video, Image to Video, Video to Video, Lip Sync, etc.
The interface is organized by model: Gen-3 Alpha, Gen-3 Alpha Turbo, Gen-4, Aleph, and variants. Each has different credit costs — Gen-4 is the most expensive but most realistic.
Core Models and Tools
1. Gen-3 Alpha and Gen-3 Alpha Turbo
Released in 2024, Gen-3 Alpha was Runway’s first model with quality close to stock footage. Accepts up to 10 seconds per generation and works in three modes:
- Text-to-Video: text prompt generates the video from scratch
- Image-to-Video: an image as starting point + optional motion prompt
- Video-to-Video: transforms an existing video into another style
Gen-3 Alpha Turbo is a faster, cheaper version (5x fewer credits), ideal for iteration and quick variations — start there and only move to Gen-3 Alpha for the final take.
2. Gen-4 (default model in 2026)
Released in late 2025, Gen-4 is the current state of the art:
- Significantly better photographic realism
- Stronger character coherence across frames
- More natural camera movement
- Accepts up to 16 seconds per generation
This is the recommended model for commercial production. The credit cost is higher, but the raw output requires less rework.
3. Aleph (Framed Video)
Aleph is the experimental model launched in 2025 for “framed video” — generation with cinematic framing control, depth of field, and specific camera movement (dolly, pan, crane, push-in). Instead of vague prompts, you explicitly define the shot type.
Useful for directors looking to reproduce a specific visual language, with close-ups, reverse shots, and full scene coverage.
4. Lip Sync
Lip Sync is one of the most underused tools. You upload a video (or animated image) and an audio file, and Runway syncs the character’s lips to the spoken audio. It works well for:
- Videos with AI-generated characters speaking pre-scripted lines
- Dubbing existing videos into other languages
- Avatar animations
The combination ElevenLabs (voice) + Runway Lip Sync (video) has become standard for creators producing content in multiple languages.
5. Motion Brush
Motion Brush is the tool that feels most like “Photoshop for video.” You upload a static image, paint with the brush over the areas that should move, and define the type of motion (vertical, horizontal, zoom). Runway animates only the selected regions.
Typical cases:
- Turn a landscape photo into a scene with wind and moving clouds
- Make a character’s hair sway while the rest stays static
- Create subtle “alive” effects on studio photos
6. Traditional Video Editor
Runway is also a complete video editor. You can:
- Cut, join, and organize clips on a timeline
- Apply transitions and effects
- Run automatic rotoscoping with AI (separate person from background in seconds)
- Use Magic Mask to mask specific objects
- Apply Inpainting in video (remove unwanted objects frame by frame)
Runway Pricing in 2026
The plan structure:
| Plan | Price/month | Credits/month | Max resolution | Watermark |
|---|---|---|---|---|
| Free | $0 | 125 (one-time) | 720p | Yes |
| Standard | $15 (annual) / $18 | 625 | 1080p | No |
| Pro | $35 (annual) / $40 | 2,250 | 1080p + Aleph | No |
| Unlimited | $95 (annual) / $119 | Unlimited in Explore mode | 4K (model dependent) | No |
| Enterprise | Custom | Custom | 4K + governance | No |
Standard at $15/month is the entry for serious personal use. Unlimited at $95 is the most chosen plan among professionals — generate as much as you want without credit anxiety.
Heads up: each generation consumes different credits. A 10-second Gen-4 video can cost 100-150 credits. Do the math before subscribing.
Real Use Cases
Film and series: studios use Runway in pre-production (animated storyboards, lighting tests, scene previsualization) and in post-production for set extensions and object removal. Lionsgate signed an official partnership in 2024.
Commercials and marketing: agencies build A/B variations of commercials in hours, not weeks. Coca-Cola released the “Holidays Are Coming 2024” commercial produced with Runway, sparking heavy industry debate.
Social media: creators produce viral clips from static images using Image-to-Video + Motion Brush.
Music: full music videos generated with Gen-4 + Lip Sync, synced to music via tools like Suno. Independent music tracks have already hit TikTok trending charts.
Corporate training: institutional videos with talking avatars instead of filming actors.
Architecture and real estate: property and development previsualization animated from static renders.
For digital content creators combining video + voice + text, also worth knowing autonomous AI agents to automate part of the production flow.
Limitations and Things to Watch
Even with Gen-4, Runway has real limits that affect professional production:
- Cross-clip coherence is still a challenge. A character generated in one clip can have small differences in the next. For long projects, this requires manual editing or complex pipelines.
- Long videos require assembly. You generate at most 10-16 seconds at a time. 1-minute videos are composed of multiple chained generations.
- High credit consumption. Iteration is expensive. Professionals often run 20-50 generations before approving a scene.
- Limited granular control. Unlike DaVinci or Adobe After Effects, Runway does not allow frame-by-frame editing with full control.
- Hands and text still problematic. As with other generative models, hands and written text inside generated scenes often come out distorted.
- Copyright in a gray area. Training models on protected material continues to be the subject of legal disputes. Assess risk before commercial use in specific markets.
Tips for Professional Results
A few practices that separate amateur video from publishable production:
Use Image-to-Video whenever possible. Pure text produces unpredictable results. Starting with an image (can be generated in Midjourney or DALL-E) gives much more visual control.
Start with Turbo, finish on Gen-4. Initial iterations on Gen-3 Alpha Turbo (cheaper and faster), final version on Gen-4.
Short, specific prompts. “Cinematic shot of woman walking in Tokyo street, neon lights, slow dolly forward” works better than a descriptive paragraph.
Use Motion Brush for subtlety. Images with partial motion (hair, water, fire) look more natural than images with everything moving.
Combine models. Use Gen-4 for hero shots, Gen-3 Turbo for coverage, Lip Sync for dialogue, Motion Brush for details.
Edit inside Runway’s own editor. The integrated timeline lets you cut, sequence, and refine everything without exporting to Premiere/DaVinci on every test.
Render in 1080p when possible. 720p screams AI origin. 1080p (Standard or above) already looks like professional production.
FAQ
Is Runway ML free?
There is a Free plan with 125 one-time credits, only enough to experiment. For real use, the Standard plan at $15/month (annual) is the minimum recommended.
Can I use Runway videos commercially?
Yes, on any paid plan (Standard or higher). The Free plan has a watermark and prohibits commercial use.
How long does it take to generate a video on Runway?
5-second videos on Gen-3 Alpha Turbo take ~30 seconds. Gen-4 videos can take 2-5 minutes. Aleph and long clips can take up to 10 minutes of processing.
Is Runway better than Sora or Veo?
In 2026 the three models compete on quality. Sora (OpenAI) has a slight edge on complex prompts and wide scenes. Veo (Google) integrates better with Workspace. Runway leads on professional tools (Motion Brush, Lip Sync, editing integration).
Can I train Runway on my own face/style?
On the Enterprise plan, yes. There are Custom Models that allow training with proprietary material to keep character or visual style consistent. On individual plans, this feature is limited.
Official sources
For deeper context, see the official sources and authoritative references below:
