The complete path: blank project → written story → framed shots → performance → a published, narrated slideshow. Twenty minutes the first time; much faster after.
1Start a project
Open Story Cards. Click the black spine on the left to open the project drawer, then ＋ (top-left).
The Project Manifest opens automatically. Fill in what you know — title, subtitle, logline, theme — and rename the placeholder characters (these become selectable speakers everywhere). Add locations; they'll appear as one-tap suggestions later. Set a cover image — it backgrounds your project's spine and the shelf card. All of it is editable later via the book icon → Project manifest…
Preferences live behind the gear on the spine: light/dark, vertical/horizontal layout, Overview (storyboard look, image backgrounds) vs Focus (clean writing board), accent color, reading pace (wpm).
2Write the cards
Each card is a scene: number → title → duration & word count → scene action → Script → On-screen.
Type the title, then the one-line scene action (what happens in frame).
Click ✎ edit script to open the script drawer. ＋ add line offers your cast (with avatars), the Narrator, or a Soundscape note (music/SFX — displayed in italics, never read aloud). Click a line's name to reassign its speaker; characters you use appear as avatars under the title.
Duration is automatic from word count at your wpm. Drag the seconds number to override; scroll it to nudge.
Hover a card's top edge for tools: drag-reorder, AI draft, lock, perform from here, dictate (mic), delete. Hover the seam between cards and click the column to insert a scene. Click a card's number to collapse it.
3Give every card an image
In the On-screen section click choose image. Type what you want — your words become the image's label — and photos appear as you type (real photography via Pexels; the photographer is credited automatically). Click a photo and it's chosen — no confirm step.
Paste an image URL to use it directly, click the upload button, or drag & drop a file anywhere on the picker. Images already in your project show first so you can reuse them.
In Overview mode the image becomes the card's backdrop — your deck reads as a storyboard.
4Cut a beat across cards
A beat that needs several images is just several cards: his face → the empty bench → his face again, closer. One image per card keeps the deck honest — a card is a beat.
Insert cards for the beat (hover the seam between two cards and click the column), then give each its own image and line.
Set the follow-on cards to cut — the small toggle next to view / edit flips a card between slide (carousel, the default) and cut (hard edit). Cuts are what make a return shot land.
For the return, open the first card's image → Frame · stickers… → Frame → drag to the focal point, zoom in → Reframe onto a new card. Same photo, tighter framing, set to cut automatically — no second copy of the image.
Opening a project made before v17? A card holding several shots offers split N shots into cards — one click converts it, moving each shot's micro-VO into that card's script.
5Motion and marks
In the Frame modal, pick the image's motion: Push in (lands exactly on your framing), Pull out, Pan, or Still. Global Ken Burns on/off is in preferences.
Stickers tab: drop emoji or type text (broadcast-style captions); drag to place, select to resize.
Draw tab: circle a detail in six colors — strokes are vector and stay crisp everywhere.
The image editor's Text overlay field prints bold type on the image itself (The 6:40 uses it for "TUESDAY").
6Perform it
Click the ◉ tool on any card (or book menu → Lock all → perform). A teleprompter stage plays: each card's image behind its words, with a timing bar keeping pace.
Space/→ next · ← back · P play/pause · Esc exit. Cards slide like a carousel unless you set one to cut.
7Narrate it — your own voice
List the steps and their lines: python3 tools/attach-audio.py stories/<slug>/index.html --list(this runs on the exported file — see step 9)
Record while you perform. Perform mode is your teleprompter — start a recorder (Voice Memos, or QuickTime → File → New Audio Recording), play the story, and read one take per step.
Name the files step-00.m4a, step-01.m4a… to match the list, and drop them in the story's audio/ folder.
Attach and redeploy: python3 tools/attach-audio.py stories/<slug>/index.html
Steps without a recording fall back to device TTS — you can voice just the title card and let TTS carry the rest, or silence a step by leaving it out and muting 🔊.
8Narrate it — Google TTS
Prefer a generated narrator? The 🔊 button normally uses the viewer's device voice — quality varies. Instead, pre-generate the audio once with Google's Gemini TTS and bake it into the story; the player automatically prefers real audio. (Runs on the exported file — see step 9.)
export GEMINI_API_KEY="your-key"
python3 tools/narrate.py stories/the-640/index.html
# pick a different voice:
python3 tools/narrate.py stories/the-640/index.html --voice Charon
It writes audio/step-NN.wav next to the story and patches each step. Redeploy and every viewer hears the same narrator, even offline.
Cost ≈ $0.015 per minute of audio — about a cent for The 6:40. Voices: Kore, Charon, Puck, Fenrir, Aoede.
9Export
Book icon → three exports:
Export slideshow (.html) — the publishable artifact. One self-contained file: click-through, auto-read (🔊), autoplay (▶), photo credits end card. Send it to anyone.
Export document (.doc) — a formatted screenplay for Word/Pages/Docs.
Export project JSON (for AI) — tight scene instructions for a video-generation model; also re-importable.
Save to file… is your backup; Open backup file… restores it (that's also how you open The 6:40's project file).
10Publish to the shelf
Export the slideshow; make a folder stories/<your-slug>/ in the shelf repo and save it as index.html.
Copy the OG <meta> block from The 6:40's file into yours (swap title, logline, cover) so links unfurl nicely.
Recorded or generated narration? Run the step-7/8 tool on this exported file now.
Copy and edit meta.json; add a card for it on the gallery page.