Generations

Type a prompt and get a video or image back, no camera or filming required.

Generations uses AI models (Sora and Kling for video, Flux and Nano Banana for images) to create content from text descriptions. B-roll, product visualizations, lifestyle clips, or standalone content. Use it to fill gaps in your content library or create things you cannot film.

Video Generation

Prompts

Describe the scene. Be specific about the subject, action, setting, camera angle, and mood. The AI is literal. More detail produces better results.

Good prompts:

• "A woman in her 50s foam rolling her lower back on a yoga mat in a bright living room, natural lighting, slow camera push-in"
• "Close-up of hands applying moisturizer to skin, soft studio lighting, shallow depth of field"
• "A phone screen showing a wellness app dashboard, slight camera drift, warm tones"

Weak prompts:

• "foam roller video"
• "make something cool"

Settings

• Duration. Depends on the model: Sora supports 4, 8, or 12 seconds; Kling supports 3 to 15 seconds. Short clips for hooks and transitions, longer for demos and sequences.
• Resolution. 720p for drafts and testing. 1080p for final content.
• Model. Sora 2 and Sora 2 Pro, Kling 2.5 Turbo, Kling 3.0 and 3.0 Pro, Kling 2.6 Motion Control, Seedance 2.0, and Lip Sync. Start with a fast model to iterate, switch to a Pro model for finals.
• Aspect Ratio. 9:16 (vertical, TikTok), 1:1 (square), or 16:9 (landscape).

@Mentions

Type @ followed by an account username to inject that account's personality into the prompt. If the account has a personality defined, the AI uses it to shape the output. The generated content matches the persona of that specific account.

Image Generation

Same prompt principles as video. Be specific about subject, composition, lighting, and style.

• Model. Flux Dev or Nano Banana. Both fast and high quality.
• Aspect Ratio. 9:16, 1:1, or 16:9.

Image-to-Video

Attach an image to your prompt to use it as a starting frame. The AI generates a video that starts from your image and animates from there. Useful for bringing still product shots to life.

The Editor

After generating a video, click into it to open the editor. The editor lets you:

• Add text overlays. Drag text blocks onto the video canvas. Style them with fonts, colors, outlines.
• Add image overlays. Layer product images or logos on top of the video.
• Trim and clip. Adjust the start and end points. Split into multiple clips on a timeline.
• Set captions. Write the post caption directly in the editor.
• Pick accounts. Assign the video to specific accounts.
• Attach music. Add a TikTok or Instagram audio track.
• Queue for posting. Send directly to the posting pipeline when done.
• Save as draft. Save your work to finish later.

The editor turns a raw AI generation into a finished, post-ready video without leaving Doublespeed.

Managing Generations

All generated content shows as cards on the page. Each card shows the prompt, a preview, generation status (pending, processing, complete), and the model used.

Filter by:

• Type. All, Video, or Image.
• Account. If you used @mentions, see generations for specific personas.

Download any generation, add it to a post, or open it in the editor.

Workflow: Generating B-Roll Without Filming

1. Identify gaps. You have 5 hook clips but need supporting b-roll. Lifestyle shots, close-ups, ambient scenes.
2. Set mode to Video. Duration 4 seconds, resolution 1080p, aspect ratio 9:16.
3. Write 5 prompts. One for each type of b-roll:
   • "Hands unboxing a small package on a wooden table, natural daylight, close-up"
   • "Morning routine, person stretching in bed, soft warm lighting"
   • "Overhead shot of a desk with a journal, coffee, and a phone, minimal aesthetic"
   • "Person walking through a farmer's market, handheld camera feel, sunny day"
   • "Close-up of steam rising from a coffee cup, slow motion, shallow depth of field"
4. Generate each one. 30 to 90 seconds per generation depending on model and duration.
5. Review. Some will be great. Regenerate the weak ones with tweaked prompts.
6. Use in the Video tool. Download the best clips, upload them as body clips in the Video tool, pair with hooks, render.

Best Practices

Prompt Quality

• Reference real cinematography: "handheld camera feel," "slow push-in," "shallow depth of field."
• Include subject, action, setting, camera angle, lighting, and mood. Every detail you add improves the result.
• Avoid abstract or vague prompts.

Model Selection

• Sora 2 / Kling 2.5 Turbo for quick iteration and testing. Faster turnaround.
• Sora 2 Pro / Kling 3.0 Pro for final content. Better motion, higher fidelity.
• Kling 2.6 Motion Control when you need precise control over movement.
• Lip Sync for talking-head style content.
• Flux Dev / Nano Banana for images. Fast and high quality.

Duration

• 4s for hooks, transitions, reaction clips.
• 8s for short demos, lifestyle b-roll.
• 12s for longer sequences. Check that motion stays coherent through the full clip.

Integration

• Generations work best as raw material. Generate clips here, combine and overlay text in the Video tool.
• Use generated images as carousel backgrounds or collection assets.
• Mix AI and filmed footage for variety without scaling production costs.

Credits

Generations spend credits: about 6 credits per video (varies by model and length) and 1 credit per image. Costs show before you confirm. See Credits for plans and balances.