From Prompt to Production: Using the AI Author

You do not need to learn the .ws language to make your first video. The AI Author takes a natural language description and generates a complete script — scenes, text, intents, timing, and all. This guide shows how to use it and how to refine the output.

Two Ways to Start

1. The Empty Canvas Shortcut

When the canvas is empty, you will see:

Type / to begin — or press Space for AI assistance.

Press Space. A copilot block appears with a text area and a Generate button. Type your prompt:

A 20-second product launch video for Apex, a project management tool.
Highlight the problem of wasted time, show key stats ($42M revenue, 10x faster),
and end with a call to action: "Start free at apex.dev"

Click Generate. The AI Author calls the server, which sends your prompt to an LLM with Workshop-specific instructions. The response is a fully formed .ws script that gets parsed back into blocks and inserted onto your canvas.

2. The Copilot Block

You can also insert a copilot block inside an existing scene. Press / in any insert point and choose AI Copilot (marked with a ✨ sparkle icon). This lets you generate content for a specific scene rather than an entire film.

Use this when you already have a structure but want AI help filling in one section:

Add a proof scene with 3 stats about market growth,
using big numbers and short labels

What the AI Author Generates

A typical AI Author response includes:

• Film block with mood, aspect ratio, and tempo
• Multiple scenes with titles and intent settings
• Text elements with size, color, emphasis, and entrance animations
• Number elements with countFrom values and labels
• Timing — beats between elements, duration hints on scenes
• Background settings — alternating dark/light, accent backgrounds for CTAs

The generated script is syntactically valid .ws and compiles immediately. You can preview it the moment it lands on your canvas.

Refining the Output

AI-generated blocks are marked with a subtle visual indicator. They are regular blocks — you can edit, reorder, delete, or reconfigure them exactly like manually created content.

Common Edits After Generation

Swap the mood. The AI picks a mood based on your prompt, but you might want a different feel. Click the Film block's mood dropdown and change it. Preview to see the difference.

Tighten the text. AI-generated copy tends to be slightly verbose. Trim lines to be punchier — shorter text generally renders with more visual impact.

Adjust emphasis. If the AI set emphasis: "powerful" but you want emphasis: "you", click the setting and change it. Emphasis drives which word gets accent color treatment.

Reorder scenes. Drag scenes by their handle (≡) to change the narrative flow. Intent auto-inference will update when you move a CTA scene from the middle to the end.

Add assets. The AI generates text-only scripts. Add image, video, or logo blocks manually and upload your media files.

Change timing. Click beat durations to edit them inline. Add or remove beats to control pacing.

Prompt Tips

The AI Author responds to structure and specificity in your prompt. Here are patterns that produce better output:

Be Specific About Content

Weak:

Make a video about our product.

Strong:

A 15-second vertical video for Instagram introducing Apex,
a project management tool for remote teams. Open with the pain
of scattered communication, show that Apex unifies chat/tasks/docs,
and end with "Try free at apex.dev".

Mention Numbers and Stats

The AI Author generates show number elements when you include data:

Include these stats: $42M ARR, 10,000+ teams, 99.9% uptime, 4.8 star rating.

Specify Mood and Aspect

You can guide the visual treatment:

Mood should be dramatic and cinematic. 9:16 vertical format.
Tempo deliberate — this is a prestige brand.

Describe Scene Structure

Three scenes: first show the problem (slow, dark), second reveal
the solution (bright, fast), third call to action (accent background).

Reference Visual Techniques

The AI understands Workshop-specific concepts:

Use a split layout for the comparison scene. Put the old way on the
left and the new way on the right. Use hero-sized numbers for the
proof scene.

The Generate Flow

Here is what happens when you click Generate:

1. Your prompt is sent to the server along with the current mood, aspect, and duration settings.
2. The server uses AI to generate a complete .ws script based on your description.
3. The generated script is validated to ensure it compiles cleanly.
4. The Studio receives the result, parses it into blocks, and inserts them onto your canvas.
5. The copilot block is replaced by the generated content.

Generation typically takes a few seconds. Longer or more detailed prompts may take slightly longer.

Workshop Studio also generates prompts for Video Language Models (Sora, Veo, Kling, Runway). This is a separate feature from the AI Author — it takes an existing .ws script and produces per-scene descriptions optimized for a specific VLM platform.

How to Use It

1. Write or generate your .ws script.
2. Click Prompts in the toolbar.
3. A platform picker appears. Select your target: Generic, Sora, Veo, Kling, or Runway.
4. The server analyzes each scene and generates a visual description tuned for that platform's prompt style.
5. A dialog shows one card per scene with the generated prompt and optional negative prompt.
6. Click Copy All to copy every prompt to your clipboard.

Use these prompts in external VLM tools to generate AI video clips, then upload the resulting files as video backgrounds in your Workshop scenes.

End-to-End Workflow

Here is the fastest path from idea to rendered video:

1. Press Space on an empty canvas.
2. Describe your video in 2–3 sentences.
3. Click Generate and wait a few seconds.
4. Preview — click the Preview button to verify the structure.
5. Edit — tighten text, swap mood, adjust timing.
6. Upload assets — add logo and background video if needed.
7. Render — click Make Video and download the MP4.

Total time from blank canvas to finished video: just a few minutes.