GA StudioUser Guide

Your Studio projects

• New project (top right) starts a new video.
• Open continues a project exactly where you left off; Delete removes it.
• Each card shows scenes, pictures, videos and the aspect ratio — 9:16 is a vertical Short, 16:9 widescreen.
• Filter by Recent, With images or With videos. Press + beside Categories to make folders, then file a project from the dropdown on its card.

Studio saves automatically after every change, so there is no Save button to remember in the wizard.

Starting a new project

Press New project, give it a name, then choose:

• Start with AI → — the recommended path. Opens the seven-step wizard at Step 1.
• Just create project — an empty project in the timeline editor, for when you already have your own pictures and script.

The Draft screen

Story — the box that decides your results

• Story or topic — a single sentence works; a paragraph with the facts, the angle and the ending you want works much better.
• Narration language — the language of the voice and the captions (separate from the Studio interface language).
• Video format — 9:16 Short / Reel for vertical phone video, or 16:9 for widescreen.
• Length — Auto lets the writer choose (30 seconds to 3 minutes for a Short), or pick 1 min, 3 min, 5 min or Manual.
• Scenes — Auto or Manual. The line underneath estimates the result, e.g. "About 1:00 · 7 scenes of about 9 seconds".

Look — how every picture is drawn

Pick one of the Presets, or switch to Manual and describe your own style.

Why some looks make two pictures per scene

The four explaining looks tell each scene in two beats. The scene opens on a clean picture while the narrator sets it up; then, at the moment the narrator gets to the point, an explained version of the same picture fades in over it.

• The pair is called Plain / Explained (Infographic and Craft), Before / After, or Outside / X-ray.
• The explained picture is drawn from its own plain picture, so it keeps the same camera, objects and colours — and can only be made after the plain one exists.
• Because each scene has two beats, these looks plan fewer, longer scenes of about 8½ seconds.
• They make still pictures only.

Characters

• Auto — the AI decides who appears and writes a character sheet for each recurring person or mascot. A topic without recurring people gets none.
• Manual — name the characters yourself.
• Make each character sheet first in Google Flow and attach it as a reference image… — keep this ticked so a character looks the same in every scene.

AI writer — who writes the plan, and when to use each

How "Write it with Gemini in Chrome" works

1. Press Write it with Gemini in Chrome.
2. First time only — you must allow access. Chrome opens a page titled Let GA Studio write with Gemini. Press Allow Gemini, then press Allow in the permission box Chrome shows. If you press Not now or Deny, Gemini cannot write the plan.
3. Chrome opens Gemini and Studio's prompt is sent there. Keep that tab visible until Gemini has finished.
4. The reply comes back to Studio and opens as your plan.
5. If anything goes wrong, Studio says so and offers the copy-and-paste route — you never lose the prompt.

⚠ The first time you use Gemini in Chrome, you have to press Allow. Chrome asks once whether Google Flow Automator may use gemini.google.com — press Allow Gemini on the page that opens, then Allow in Chrome's box. Without it, nothing is written. Missed it? Press Write it with Gemini in Chrome again and the question comes back.

💡 This route is available for Gemini only. For ChatGPT or Claude, connect an API key or use the copy-and-paste route.

The Plan screen

Story spine — four lines that hold the video together

• Why watch — why a stranger would keep watching.
• Hook — the exact first sentence, meant to grab attention in 3 seconds. Sample: "Two billion people depend on mountain ice."
• Turn — what changes in the middle.
• Ending — the exact last sentence, so the video lands instead of trailing off.

Edit any of them directly.

Scene plan

The header shows the totals (e.g. "7 scenes · 1:03 · Craft infographic"). Each scene card lists the Image prompt (what Google Flow will draw), the Explained picture (for two-picture looks), and the Narration. Toggle a character name to choose which character sheets are used as references.

• Edit — change one scene.
• Rewrite the plan — have the writer produce a new version.
• Add character — add a character sheet yourself.
• Make the images → — continue to Step 3.

Before you press Send

• Chrome is open and Google Flow Automator is signed in.
• You are signed in to Google Flow with the same Google account.
• Check the right-hand panel — How Flow is asked, Image model, Explained picture model and Label style. These travel with the Send, so there is nothing to set in the extension.
• "Not connected to this window: Send opens Google Flow in Chrome…" is normal. Studio simply hands the prompts over through Chrome.

What happens when you press Send to Extension

A picture takes about 1–3 minutes in Google Flow (the sample averaged 2:24), so a 7-scene two-picture video takes roughly 15–20 minutes. The extension also rests between pictures — Studio says On a break · resumes in… or Next picture in…. Keep Chrome open and the PC awake; nothing needs clicking.

The order pictures are made in

1. Character sheets first — only when the plan has characters and the reference box is ticked. Press Make the character sheets, then approve each with Use this sheet (or Make again / Upload my own). Send to Extension stays off until every sheet is approved.
2. Plain pictures — one per scene, in scene order.
3. Explained pictures — two-picture looks only, sent automatically once the plain pictures are in.

Buttons on the Images step

A picture with a problem is outlined in red with the reason under it. Press Make the N marked pictures again, or Looks fine if the picture is actually right.

Google Flow settings (right panel)

• How Flow is asked — Flow page (default): the extension works in the Google Flow page at its usual pace, and the Flow tab comes to the front. Flow API: faster, no visible tab needed; it is experimental, and Studio switches back to Flow page by itself if needed.
• Mode — Image. Video is not available yet.
• Image model — the model for plain pictures.
• Explained picture model — text inside a picture is where models differ most; Nano Banana Pro draws labels best.
• Label style — how labels look in explained pictures: UI mode or Background mode. The previews under the buttons show the difference.

If the extension needs you

The Review screen

If a picture is wrong

• Make this picture again or Make again in Flow — asks Google Flow for a new version.
• Make this explained picture again — redraws only the explained version from the current plain picture.
• Replace with a file — use your own image instead.
• Look for the new picture in downloads — if Flow made it but it never arrived.
• Open Image prompt or Explained picture prompt under the narration to see — and edit — exactly what was asked for before trying again.

Appears at — timing the explained picture

Appears at is the second at which the explained picture starts to fade in. Studio sets it to the moment the narrator says the words the explanation is about, and updates it whenever the narration changes. Type a number to set it yourself; Match the narration hands control back to Studio.

When every scene shows Picture ready (and Explained picture ready), press Looks good → Captions & overlays.

Styling captions

• Scenes (left) — pick a scene to preview, or use Previous scene / Next scene. A CC mark means the scene has captions.
• Auto / Manual — Auto picks a size, position and font for your format and language. Change anything and Studio switches to Manual; press Auto to go back.
• Font, Weight, Size, Colour, Background, Box and Position apply to every scene. You can also drag the caption on the preview and resize it by its corners.
• Overlays tab — extra layers on the picture for the scene you are on.

Edit all captions

Every caption of the video in one list, one line = one caption. Fix a typo or split a long caption onto two lines; changed words also become that scene's narration, so voice it again afterwards.

The Narration & music screen

Narration — choosing a voice

• Auto picks a voice that speaks your narration language, on this PC when possible. Manual uses the voice you choose for every scene.
• Voice service, then Voice.

• Scenes to generate — Only scenes that still need audio (normal) or Every scene with narration (after changing the voice).
• Preview voice plays a short sample; Generate narration voices the scenes.

Background music

• Auto — Local BGM writes music for your story, mixed quietly under the voice and faded out at the end. Manual — set the volume and fades yourself.
• Music direction — genre, mood, tempo and instruments. The writer suggests one from your story.
• Install Local BGM — needed to compose music on this PC (about 3 GB; accept the Stability AI Community License first).
• Upload a track — use your own music.
• No BGM: make this video without background music — the step is not finished until you have music or tick this.

Make narration and music → voices anything still missing, makes the music if needed, and opens the last step.

Preview

Press Play. The ticks under the player — Pictures, Narration, Subtitles — confirm nothing is missing. Change narration or music jumps back to Step 6. Two-picture looks keep their pictures still so the explained picture lines up exactly with the plain one; other looks add gentle photo motion.

Export options

Resolution: 2K, 1080p (recommended), 720p or 480p. Quality: High quality, Balanced or Smaller file. Finished videos are saved in the renders folder inside the project folder.

The fast route — One Click with Gemini

One Click with Gemini on the Draft step runs Steps 2 to 6 for you: Gemini writes the plan, Google Flow draws every picture (character sheets are approved automatically), a voice on your PC reads the script, and Studio stops at Preview & export. Nothing is rendered until you press Make the MP4, so check the pictures on Images and Review first.

When to open the editor

Open the editor (bottom of every wizard step) switches to a scene-by-scene timeline; Story steps returns to the wizard.

• Scene tabs on the right: Scene (name, duration, prompt, replace picture), Audio (narration, speed, voice, CC Retiming), Subtitles (start and end of each caption), Overlay (text, stickers, emoji, images, templates) and BGM.
• The timeline has rows for Visuals, Motion, Narration, Subtitles, Overlay and BGM. Drag clip edges to change durations; Split and Delete act on the selected clip.
• AI → Narration / Caption Editor shows the whole script in one dialog, with Save subtitles as CSV and Import edited CSV. AI → Generate Approved Narrations voices every approved scene at once.
• The Project menu holds Project settings, Export / Import prompt CSV, Export, Save and Project List.

The Settings tabs

Local AI & Models

Local AI comes in Medium and High. Medium needs a dedicated graphics card with at least 6 GB of memory; High needs more. If your PC is below that, skip Local AI and use Gemini in Chrome — it writes better plans anyway. Run the System Check if you are not sure.

Do I need API keys?

No. Gemini in Chrome, the copy-and-paste route and the local services are free. Add your own keys for cloud writers and voices — that spends your own credits with those providers. All keys are stored locally on your PC.

Send to Extension does nothing

Open Chrome, make sure Google Flow Automator is signed in, and press Send to Extension again. See If the extension needs you for the message Studio shows.

Gemini in Chrome does not finish

The first time, make sure you pressed Allow Gemini and then Allow when Chrome asked — press Write it with Gemini in Chrome again if you missed it. Keep the Gemini tab visible until it is done, and check you are signed in to Gemini with your membership account. If it keeps failing, use the copy-and-paste route.

A picture does not match the narration

On Review, open Image prompt, adjust it, then press Make this picture again.

The explained picture appears too early or too late

On Review, set Appears at, or press Match the narration.

There are no voices to choose from

Install Local TTS in Settings → Local AI & Models, or connect an ElevenLabs or OpenAI key.

Narration & music will not finish

Add background music, upload a track, or tick No BGM.

I edited the narration but the voice didn't change

Changed words need new audio. On Narration & music, choose Only scenes that still need audio and press Generate narration.

Local AI is very slow or fails

Your graphics card is probably below 6 GB, or another heavy app is using it. Use Gemini in Chrome instead.

Studio won't open from the extension

Check the extension is signed in with the account that holds your membership. Trial and Starter accounts cannot open Studio.

How do I report a problem?

Use Help → Feedback & logs. Describe what you were doing and tick Attach privacy-safe diagnostic logs — they describe how Studio ran and never include your keys or media.