AI video generator: from idea to your first clip

Watch the output. Choose a direction.

Before choosing a subscription, see the kind of short footage you want to make. These are real clips from AIGE’s library. They illustrate different visual directions, rather than a controlled comparison of models.

The editable prompts below are new templates for similar ideas, not the original prompts used to make these clips. A new generation will produce a different result.

Kling, Veo or Seedance: choose the model for this shot

There is no single best model for every brief. Start with the input you have, the sound you need and your testing budget. Compare one short shot before planning a whole sequence.

Model availability depends on your plan. The cost table and calculator show example settings; the studio confirms the current choices and quote.

How to use an AI video generator: your first clip in five steps

1. Give the clip one job. Pick an opening hook, a character action, a product detail or an establishing shot. Decide where it will appear: a vertical social edit or a wider scene.
2. Draft the idea and look. The creation flow asks what you want to make, lets you choose a visual direction, and shows your written brief. Edit it until it describes one useful shot.
3. Choose your plan with the cost in view. Compare the credit allowances and model access, create your account, and complete checkout. Drafting does not spend generation credits.
4. Prepare the generation in the studio. Open your draft, attach a reference frame if needed, then choose the model, available length, aspect ratio, resolution and audio. Review the quote before pressing Generate.
5. Review, revise, then edit. Watch the whole clip. Keep a usable take or change one instruction for the next attempt. Assemble accepted shots in the video editor and export your edit.

You choose when to generate. Completing onboarding or payment opens your draft; it does not automatically create a video or spend its generation credits.

Write a shot, not a whole movie

# Your first AI video shot

## Brief
Subject + setting + one action + framing/camera + lighting/style.
If you need sound, describe it separately and select a supported audio model.
If you need a particular face or product, prepare a reference frame you have permission to use.

## Before generating
- Pick the model, duration, aspect ratio and available resolution in the studio.
- Review the actual credit quote and plan requirement.
- Start with a short test. Leave room in your budget for revisions.

## Templates
### Everyday scene
Medium full shot of a woman in a tan blazer at a city crossing, holding a coffee. She looks up and takes one step forward. Gentle camera push-in, warm late-afternoon light, natural street style. One continuous shot.

### Cinematic world
Wide establishing shot of a stone castle at night, torches glowing below and streaks of fire crossing the cloudy sky. Slow forward camera movement, dark fantasy atmosphere, one continuous shot. No dialogue.

### Surreal visual
Vertical wide shot of a festival crowd at sunset beneath giant luminous jellyfish. The jellyfish drift slowly overhead as the crowd watches. Mostly still camera, soft glowing light, dreamy surreal atmosphere. One continuous shot.

These are new templates for similar ideas, not the original prompts for the library clips.

## Review
Watch the full clip. Check the face, hands, clothing, product shape, background, motion and sound.
Change one instruction at a time. Assemble accepted shots in the video editor.

## References
- AIGE model/settings guides: https://aige.ws/models/kling/ · https://aige.ws/models/veo-3/ · https://aige.ws/models/seedance/
- Google DeepMind prompt guidance: https://deepmind.google/models/veo/prompt-guide/

Working in Claude or ChatGPT? Connect AIGE over MCP and ask in plain words; your credits and saved characters come with you. Hand the card to your assistant as context. Set up the connector

Make the request specific enough to test

For the underlying prompt elements, see Google DeepMind’s official Veo prompt guide. For controls exposed here, use the AIGE model guides above.

Examples from AIGE’s default credit table, checked September 30, 2026. Different settings use different amounts; the studio’s current quote is authoritative. Plans have a credit allowance, not unlimited generations.

What AI video costs: budget for attempts, not just the finished seconds

A usable shot may take more than one generation. If you budget three attempts at the same settings, multiply the per-generation credit quote by three. For example, three 5-second Kling V3 Standard silent attempts use 126 credits at the default rate; they are three separate outputs, not a guaranteed 15-second finished video.

Test a shorter length or a less expensive available setting first. Once the composition works, decide whether extra resolution, a longer take or sound is worth the additional credits. If a particular model needs a higher plan, compare its access alongside the allowance.

Keep the character recognizable across shots

Use a reference when the face, outfit or product matters. Text alone asks the model to invent those details again. A clear start frame gives it a more concrete starting point.

1. Prepare the still. Use your saved character or an image you have permission to use. Check the face, clothing, hands and composition before animating it.
2. Describe movement, not a new identity. Ask for a small action that fits the frame. Keep the appearance and lighting description stable across related shots.
3. Check continuity before the edit. Compare the beginning, middle and end of each clip. Reject a take if the face, outfit, product or background changes in a way that breaks the sequence.

References reduce guesswork; they do not guarantee identical details in every frame. For a sequence, keep a simple shot list and compare accepted takes together.

Turn good clips into a finished video

Generation makes the footage. Editing gives it a beginning, a useful pace and an ending. For a first social edit, plan three beats: an opening visual, a detail or action, and a closing shot. Generate them separately so you can replace one weak take without rebuilding the entire sequence.

Bring accepted clips into AIGE’s video editor, arrange and trim them on the timeline, then review the export. Check text legibility, audio, framing and the destination platform’s requirements before publishing.

Creating a clip does not automatically post it to your social accounts. Keep publishing as a separate review step.

Common problems, and what to change

• The face or product drifts: improve the reference frame, simplify the motion, and review the full output. A reference is guidance, not a guarantee.
• The motion is confused: reduce competing actions or camera directions. Test one change at a time so you can tell what helped.
• Text or logos look wrong: keep essential wording and precise brand marks out of generated footage; add and verify them in the edit.
• The dialogue or sound misses the brief: shorten the request, use a supported audio setting, and listen to the complete result before publishing.
• One clip does not tell the whole story: split the idea into shots and assemble the sequence in the editor.
• The credits run out before the edit is ready: budget for revisions, start with short tests, and review each quote before another attempt.

AI video: questions before you start

Already building with an AI assistant?

You can also connect AIGE through MCP to supported assistants and automation tools. Model options can differ from the studio, and generations use credits. Start with the connector documentation rather than assuming every studio control is exposed to an agent.