Touchizen
TouchizenCreate more, work less
Back to blog
AutoFlowCutTutorialAI ImagesWorkflow

Use Any AI Image Tool with AutoFlowCut: Bring Your Own Images into Scenes

Touchizen·

"I already made my images somewhere else."

You've got your narration recorded, your SRT timed, and your scenes grouped in a CSV. But you didn't generate your images inside AutoFlowCut — you made them in Midjourney, DALL·E, Stable Diffusion, or some other tool you love. Now you're staring at a folder full of 0001.png, 0002.png, 0003.png and wondering:

"How do I get these into AutoFlowCut so each image lands on the right scene — without breaking my SRT and narration timing?"

Good news: you don't need to edit the project file, and you don't need any plugin. AutoFlowCut is built so that your image filenames are what connect images to scenes. Get the names right, drop the files in one folder, reload — done. Your SRT and narration timing are stored separately and never get touched.

Here's exactly how it works.

The one rule: filenames control everything

AutoFlowCut does not treat the paths inside project.json as the source of truth. Every time you open a project, the app re-scans the scenes/ folder and matches each file to a scene by its name.

That means the reliable way to attach your own images is simple:

Rename each image to match its scene ID, then drop it into the project's scenes/ folder.

No JSON editing. If you hand-edit an image path into project.json, the app just overwrites it on the next load — so don't bother. The filename is the contract.

Where is the project stored?

Each project is a folder, not a single file. You'll find it here:

Windows:  C:\Users\<you>\Documents\AutoFlowCut\<project-name>\
macOS:    ~/Documents/AutoFlowCut/<project-name>/

(If you picked a custom work folder inside the app, it lives under that instead.)

Inside, the structure looks like this:

<project-name>\
    ├── project.json     ← scenes, prompts, SRT track, narration timing
    ├── scenes\          ← your images go HERE
    ├── references\
    ├── videos\
    └── sfx\

The file that actually controls scene ↔ image matching is not a field in project.json — it's the filename of each file in the scenes\ folder.

The naming rule (this is the part people get wrong)

Scene IDs in AutoFlowCut are:

scene_1, scene_2, scene_3, ...

That's 1-based, with no zero-padding. So the app looks for:

scenes/scene_1.png
scenes/scene_2.png
scenes/scene_3.png

Your exported files are probably named 0001.png, 0002.png, … — those will not match. You have to rename them:

Your file Rename to
0001.png scene_1.png
0002.png scene_2.png
0003.png scene_3.png

Extension priority: if a scene already has an old file, the app picks the first match in this order — png → jpg → jpeg → webp → gif → mp4 → webm. To avoid a stale image winning, either use .png for everything or delete the old file for that scene first.

Step by step

  1. Open your project folder (e.g. …\Documents\AutoFlowCut\my-project\).
  2. Rename your images to the scene-ID pattern — scene_1.png, scene_2.png, … (not 0001.png).
  3. Copy them into the scenes\ subfolder.
  4. Reload the project in AutoFlowCut. It scans scenes/scene_N.*, auto-attaches each image to its scene, and marks it "done."

That's it. Your SRT lines and narration timing stay exactly where they were.

One thing to double-check: scene order with a grouped CSV

When you import a grouped Scene CSV, AutoFlowCut assigns scene IDs sequentially in import order — scene_1, scene_2, scene_3, … — independent of the scene number column in your CSV.

  • If your CSV's scene numbers are already 1, 2, 3… in order, everything lines up perfectly.
  • If they skip or get reordered, match your image numbering to the on-screen scene order in the app, not the raw numbers in the CSV.

Quick sanity check: open scene 1 in AutoFlowCut and confirm it's the scene you think scene_1.png belongs to.

Let GPT or Claude do the renaming

This is a great little job to hand off to an AI assistant. Instead of asking it to edit the project file (which won't stick), ask it to generate a rename script for your files:

"I have these files: 0001.png through 0240.png. Write a script that renames them to scene_1.png through scene_240.png (1-based, no zero-padding)."

Run the script, drop the results into scenes\, and reload. That's the part where AI actually saves you time.

What about video?

This is where it's important to be honest about how AutoFlowCut works today.

AutoFlowCut does not currently support importing your own video files into scenes. There's no "attach a video" option in the UI — the app assumes scene videos are produced by its own generation step and saved into the videos\ folder with internal IDs (like t2v_5.mp4 / i2v_5.mp4). Unlike images, video files are **not** blanket-scanned from disk, so simply dropping an .mp4 into a folder won't attach it to a scene.

If your goal is a video that uses clips from an external tool (Sora, Kling, Runway, etc.), the practical path is:

  1. Bring your images into AutoFlowCut using the method above.
  2. Export the project to CapCut or Premiere.
  3. In your editor, replace the still on any scene with your external video clip. Because the timeline is already built to your SRT timing, you're just swapping the media — the timing and subtitles stay intact.

That keeps your workflow clean and avoids fighting a workflow the app wasn't designed for.

The one-line takeaway

The source of truth for scene ↔ image matching is the filename scene_<N>.<ext> inside the project's scenes/ folder — not the project file. Name your images scene_1.png, scene_2.png, … drop them in scenes\, reload, and every image lands on the right scene while your SRT and narration timing stay perfectly in place.