Start here: choose your recorded voiceover. Everything else — captions, AI footage, the finished video — is made from it.

Pause gap: silences shorter than this count as part of the same phrase. Silence floor: anything quieter than this share of your loudest moment counts as silence. Raise the floor if there is background hiss.

Reel Forge listens to your voiceover and writes the captions itself, on your own computer. The first time, it downloads a speech model (about 40–250 MB depending on accuracy); after that it is remembered and starts at once.

Changes show in the preview in step 7 straight away.

Optional. A cut-out picture of the person the channel is about, shown over the footage for the whole video. A PNG with a transparent background looks best.

Every scene gets its own AI footage: real Agnes AI video clips for the scenes you choose, and AI pictures that move with a 3D camera for all the others. The visual theme is added to every AI prompt so the whole video looks like one film.

Agnes accepts one new AI video clip per minute and each takes a few minutes, so 8 clips add roughly 10–20 minutes. A scene whose clip is not ready in time gets an AI picture instead.

One row per scene. Every scene has its own picture description; edit it and press Enter for a new AI picture. “New picture” paints another picture; “AI video” makes an Agnes AI video clip for that shot.

Optional. Title, description with chapters, tags and a thumbnail, ready to paste when you upload the video yourself. Nothing is uploaded from here.

If something does not work, press “Run checks”. Each line says whether that part of your browser can do the job. If something fails, “Copy report” and send it to whoever helps you.