Use ChatGPT, Gemini, Grok and Muse

How it works

Mark Flow has no ChatGPT, Gemini, Grok or Muse account of its own. It drives that site's tab in your Chrome the way you would: open the site, paste the prompt, press send, wait for the result and download it. So the results and the limits all come from your own account on each site.

You pick the platform with the buttons (pills) in the Gen tab of the side panel (the panel on the right edge of the browser): Google Flow, ChatGPT, Grok, Gemini, Muse.

Before you start

• The Pro or Agency plan. The Free plan only covers Google Flow: the ChatGPT, Grok, Gemini and Muse pills show a crown and open the Upgrade dialog. See Plans and billing.
• You are signed in to Mark Flow in the side panel.
• You are already signed in to each site you want to use, in the same Chrome where Mark Flow is installed: chatgpt.com, gemini.google.com, grok.com, muse.ai.
• Your account on that site still has image or video generations left.

What each platform can make

Run prompts on ChatGPT, Gemini, Grok or Muse

1. Open the side panel (shortcut Alt+S) and choose the Gen tab.

2. Click the platform pill just below the prompt box: ChatGPT, Grok, Gemini or Muse. Mark Flow opens that site's tab (or switches to the one already open).

   

3. Type your prompt in the Prompt box. To run several, turn on Multi-Prompt and leave a blank line between prompts. See also Batch generation.

4. Pick the options on the settings bar: image or video (Gemini, Grok, Muse), model, aspect ratio; for Grok, image quality or video length and resolution. Add reference images if you need them.

   

5. Click Generate. The Confirm Run dialog shows the platform name. For ChatGPT, Gemini and Grok it also shows a status: Checking..., Ready, Not logged in, or Waiting Cloudflare... (Grok).

   

6. When the status says Ready (Muse shows no status), click Run. Mark Flow switches to the site's tab, pastes the prompt, sends it, waits for the result and only then moves to the next prompt.

7. Follow progress in the Logs tab. To save files to your computer, turn on Auto download in the Gen tab before running (see Auto download).

Per-platform settings

1. Click the gear icon at the top of the side panel → Settings.

2. In the left menu, choose Providers. Each platform has its own card.

   

3. Change what you need (see the list below), then click Save changes.

• ChatGPT: default Model and Ratio; Fallback prefix is text Mark Flow adds to the start of the prompt when it cannot switch ChatGPT into image mode, so ChatGPT knows you want an image; Close the tab when done; Delete the chat after generating.
• Gemini: Model, Video ratio, Close the tab when done, Delete the chat after generating.
• Grok: default mode Image or Video, Ratio, Image quality (Fast / Best), Video length, Resolution, Close the tab when done.
• To have a platform already selected each time you open the Gen tab, go to General › Default provider.

Common problems

• A pill shows a crown and opens the Upgrade dialog: the account is on the Free plan. ChatGPT, Gemini, Grok and Muse need Pro.
• The ChatGPT login required dialog (You are not logged in to ChatGPT. Open a tab to log in?): click Open tab, sign in to ChatGPT in the new tab, then click Generate again.
• The Grok login required dialog: sign in to grok.com in the same Chrome and run again.
• Not logged in in the Confirm Run dialog: switch to that site's tab, sign in and wait for the status to change to Ready.
• Waiting Cloudflare... (Grok): Grok is showing a check that you are human. Open the Grok tab, complete the check yourself, then run again if needed.
• A message like ChatGPT: Đã hết lượt tạo ảnh. Vui lòng thử lại sau. (out of image generations): your account on that site has no generations left. Wait, or upgrade your account on that site.
• Nội dung bị chặn do vi phạm chính sách. (content blocked by policy): the site refused the prompt. Edit the prompt and run again.
• Duplicate tabs detected: you have several tabs of the same site open. Click Close extra tabs to let Mark Flow close the extras.

For other errors, see Troubleshooting.