# Chat Octopus vs Google Flow

Chat Octopus edits the recording you already have, or generates new footage, in one thread. Google Flow generates clips from a prompt or a photo.

Attach your recording to a chat and say what needs to happen to it: a caption fixed, an object gone from the shot, three clips cut out of it. The result comes back in the same thread.

## What can you do with the footage you already have?

Ask for three vertical clips, captions timed to the word, or those same words as [a separate SRT file](/guides/captions-vs-subtitles), the caption file every player and site accepts. A touch-up on the shot itself works the same way: [relight a dim take or swap out the background](/tools/video-editor), and the result is your actual footage with one thing changed, not a new clip that looks similar.

The touch-up runs on an MP4 or MOV under 200 MB, three to ten seconds per request, so trim the source down to the exact stretch first.

## What if the shot doesn't exist yet?

When there's nothing to upload, [describe the shot here](/tools/ai-video-generator) and it comes back four to twelve seconds long with sound already on it: effects, ambience, speech if you asked for any. Ask for it silent and it comes back silent. You pick from seven shapes, so a shot meant for a phone screen arrives tall rather than cropped down from a wide one. It renders at 720p, or 480p if you say so, and twelve seconds is plenty for [the short shots an edit cuts away to](/guides/what-is-b-roll), which run two to four seconds.

A photo can go in with the description too. Up to four reference pictures, people included (though not for accounts in the EEA, Switzerland or the UK), hold a face or a product steady while the clip moves. That clip runs three to ten seconds, wide or vertical only.

## What can Flow do that Chat Octopus can't?

A single generated shot here tops out at twelve seconds. Past that, a longer sequence has to be built from separate clips instead of one unbroken take, and that's where Flow offers what this doesn't: a scene builder that carries on the shot it already made, with on-screen camera controls.

Ask for a thirty-second sequence here and it generates the first clip, uses its last frame as the starting picture of the second, and joins the finished clips at the end. Each clip has to finish before the next can begin, so a thirty-second sequence is three or four clips generated one after another. When the shots are separate but need one consistent look, the same photo of a character goes into every clip, and those can all generate at the same time.

There's no camera-control screen here. Ask for a slow push in as part of the prompt.

## Do you need a special sign-in for the AI video tools?

Chat Octopus looks at how you signed in. An account made with Apple or Google can use the AI video tools, generating from a prompt, generating from a photo, or touching up footage you already have. One made with only an email and password after 28 June 2026 can't, though cutting, captioning, and reformatting work either way. An account opened before that date keeps full access.

Flow's question is whether you're paying. Generation needs a Google AI Pro subscription, which includes 100 generations a month. On Google AI Ultra the limit is higher and new releases arrive early.

## What should you try first?

Generate the shot you never filmed, then ask for it cut into the recording you did, with captions across the whole thing. Both files sit in one chat, and the edit comes back finished.