Agnes Video Generator

Manual Mode Guide

Pause at generation checkpoints to review or refine artifacts before continuing — auto for vibe, manual for precise output.

Agnes Video Generator now offers a Manual Mode: the default Auto Mode runs the whole pipeline in one pass, great for "give me a rough idea and vibe-generate". Manual Mode pauses at the checkpoints you choose, letting you confirm, modify or regenerate intermediate artifacts before continuing — ideal for creative work with a clear target output, and you can hand artifacts to stronger external AI or other tools to enhance the result.

Auto Mode vs Manual Mode

Both modes share the same pipeline — the only difference is whether you stop to inspect along the way:

Auto Mode

vibe video generating

Enter a creative idea and the whole pipeline runs in one pass: storyboard → references → video clips → narration → subtitles → final cut, with no intervention.

  • ✓Best for vibe video generating from a rough idea
  • ✓Zero manual work, one-click output
  • ✓Some randomness in results — great for exploring ideas

Manual Mode

precise output

Pause at checkpoints to confirm or refine each artifact before continuing. Tune the storyboard, polish the narration, fix subtitles, refine clips — or hand artifacts to stronger external AI or tools and feed the result back in.

  • ✓Best for clearly specified output expectations
  • ✓Every stage can be previewed, modified or regenerated
  • ✓Works with stronger external AI / ffmpeg / PIL tools

Where Manual Mode pauses

Pause points differ by task type: creative long videos support fine-grained checkpoints — every stage that produces an artifact can pause; manuscript, poetry and talking-head tasks use the standard checkpoints. Tick the points you want when creating a task:

Fine-grained checkpoints (Creative)

  1. 1

    Image analysis

    Analyze reference and end-frame images

  2. 2

    Story

    Generate the story plot

  3. 3

    Script

    Generate the storyboard script and narration

  4. 4

    Character reference

    Generate the character reference image

  5. 5

    End-frame prompts

    Generate end-frame prompts for each scene

  6. 6

    End-frame images

    Pre-generate end-frame images

  7. 7

    Video clips

    Preview and refine each scene’s video clip

  8. 8

    Narration

    Listen to the voice-over; polish the script before re-synthesis

  9. 9

    Subtitles

    Check subtitle timing and line breaks

  10. 10

    Final cut

    Confirm the final composited video

Standard checkpoints (Manuscript / Poetry / Talking-head)

  1. 1

    Storyboard

    Review or refine the script, visual descriptions and narration text

  2. 2

    References

    Check the character reference image (talking-head)

  3. 3

    Video clips

    Preview and refine each scene’s video clip

  4. 4

    Narration

    Listen to the voice-over; polish the script before re-synthesis

  5. 5

    Subtitles

    Check subtitle timing and line breaks

  6. 6

    Final cut

    Confirm the final composited video

Default pause points

Default pause points are pre-filled by task type when creating a task (add or remove any; unchecked ones pass automatically):

  • →Creative: story, script, character reference, video clips, subtitles
  • →Manuscript / Poetry: storyboard, video clips, subtitles
  • →Talking-head: storyboard, video clips

Where manual mode is not supported

Simple (single-shot) text-to-video and simple image types do not support manual mode (artifacts are only listed after completion).

Four ways to handle an artifact

When paused at a checkpoint, pick any one way to handle the current artifact:

🤖 AI modify for me

Type a change request (e.g. "make scene 2 more cinematic"), the system modifies it via model, review the diff and apply.

✏️ Edit myself

Copy the artifact path, edit it locally (JSON / subtitles / images / video), then click "modified, continue".

🤝 External Agent

Copy the collaboration prompt to a stronger local Agent (e.g. opencode / CodeBuddy CLI), then feed the result back and continue.

More free AI tools →

💻 Online editor

Edit text artifacts (script / narration / subtitles) directly in an in-page editor, compare with the original anytime, save and continue.

Only affected steps re-run

After a modification, the system re-runs precisely by the artifact-level dependency graph: editing narration re-runs only audio/subtitles/final; editing visual descriptions re-runs only references/videos/final. Unaffected artifacts are kept as-is.

Switch anytime while running

An auto task can switch to manual in one click (pauses at the next safe point); a paused manual task can switch back to auto (clears pause points and runs to completion immediately).

How to choose: auto or manual?

There is no single right answer — pick what fits:

  • →Quick output & exploring ideas → Auto Mode, vibe video generating
  • →Clear visual / style / voice expectations → Manual Mode, fine-tune each artifact
  • →Want output beyond the model default → Manual Mode + stronger external AI or tools

Start creating

The online Demo uses Auto Mode; Manual Mode is a desktop feature — download and choose "Manual" when creating a task to experience checkpoint pauses.

Open online DemoDownload desktop

Related Pages

Quick Start: Free AI Video Generation Guide

From getting a free API key to generating your first AI video — a step-by-step guide to Agnes Video Generator.

Agnes Self-Hosting Guide — Docker, npm, AI Agent | Free AI Video Guide

Run the free open-source AI video generator on your own computer or server: start.sh manual deployment, one-command Docker, npm CLI, AI Agent — complete commands and requirements.

Narration & Subtitles: Add Voice-over and Captions

The digital human mode ships with built-in TTS and lip-sync; for regular text-to-video, add captions in an editor.

Project & Documentation Basis

The checkpoint and smart-re-run mechanics described here follow the project's official repository and API documentation.

  1. agnes-video-generator — 开源 AI 视频生成项目lcy362 · GitHub (2026)

    The manual/auto mode implementation and usage follow the open-source repo's README and source code.

  2. Agnes AI Open API — 视频 / 图片生成接口Agnes AI · apihub.agnes-ai.com (2026)

    The generation endpoints behind each checkpoint (video/image/narration/subtitles) are specified in the Agnes Open API docs.

  3. opencode — 开源 AI 编程助手SST / opencode 社区 · opencode.ai (2025)

    opencode, recommended for the external-agent channel, is open source; its capabilities follow its official docs.

Why you can trust this

Written by a practicing developer and cross-checked against authoritative primary sources such as arXiv papers, vendor technical reports, and the Stanford AI Index.

S

SandGrid@lcy362

Author of Agnes Video Generator · Full-stack Developer

Independent developer and author of Agnes Video Generator, an open-source (MIT) AI video generation tool built on Agnes AI's free video models. Has helped 1,000+ creators produce AI videos at zero cost. Focused on making video generation models accessible and production-ready; content is based on hands-on practice and primary research.

View on GitHub

Last updated：2026-08-21

Ready to Start Creating?

"Making world-class AI belong to everyone." — Bruce Yang. It's completely free, no credit card, and you won't need a high-end GPU. Your first AI video starts at zero cost. Want to use Agnes AI's free video models? This is the easiest way in.

Try Online DemoGet Started Free

Clone the GitHub repo and launch in 2 minutes