One fixed image for the whole batch
One fixed image (or none) for every prompt in the run. For series sharing the same visual reference on every generation.
Useful for: keeping a consistent look without switching references between clips.
Side panel with cross-AI capture (ChatGPT, Claude, Gemini, Grok), Normal and Sequences modes, auto-download straight from Google's infrastructure. Start free, no card.
Flow has a clear, predictable workflow: write the prompt, pick the Veo model in the dropdown, Veo processes it for 1–2 minutes, download the clip, start over. Veo's quality isn't the problem — the problem is your role gets reduced to operating the mouse.
If your series has 30 scenes, that's 30 manual cycles. If you're producing a faceless channel with one daily episode of 10 clips, that's 70 cycles a week. Flow AutoKuak breaks that loop — including the image re-selection and the auto-selection of the correct Veo model.
Flow AutoKuak is a Manifest V3 extension that automatically recognizes any open tab on labs.google/fx/{user_id}/tools/flow/... (the real Google Labs path with your user ID) and opens the bridge-panel with a 4-step guided flow: source → prompt capture → preview → Groups. The ⚡ Google Veo / Flow chip in the panel header reminds you which suite tool you're using.
Pick where prompts come from. Source tab (Google Doc, Notion, HTML) or a conversation in another AI: ChatGPT, Claude, Gemini, Grok, or X.
Set a start delimiter and an end delimiter, choose the range, and the extension extracts each fragment as an individual prompt.
Review the full list before firing. Edit, discard, or reorder. In this step you also pick which Veo or Nano Banana model to use — Flow AutoKuak will auto-select it in Flow's native dropdown before every prompt.
Group prompts by batch or label to process them in blocks, with downloads organized per group.
When you hit Start, the screen splits automatically with Google Flow on the left and the Flow AutoKuak side panel on the right. Real-time logs show which prompt is processing, which Veo model is active, elapsed time, and message status. The extension captures the download URLs via chrome.webRequest and passes them to chrome.downloads to save the videos in your chosen subfolder. Technical permissions Flow AutoKuak uses (declared in its manifest.json): webRequest to detect generated video URLs, downloads to save them to your local folder, and debugger to guarantee reliable interactions with frequently-changing Flow DOM elements. No permission is used outside Flow's domain.
Flow exposes several models depending on your active plan in Google Labs. Flow AutoKuak auto-detects them and lets you pick which one to use for each batch from the side panel.
Veo 3.1
The main video model, maximum cinematic quality. 1–2 minutes per short clip.
Veo 3.1 Fast
Faster variant with lower quality, useful to iterate storyboards before launching the final batch on Veo 3.1.
Veo 3.1 Lite
Ultra-fast variant for previews at minimum cost.
Nano Banana Pro
High-quality image generation for storyboards prior to video.
Nano Banana 2
Faster Nano Banana variant, useful to iterate visual references quickly.
Before every prompt, Flow AutoKuak opens the native Flow dropdown and selects the model you've chosen. If Flow switches the model between prompts (because of Google quota restrictions, for example), the extension forces it back before sending. This kills one of the most common manual-batch errors: discovering at the end that half your batch ran on Veo Fast instead of Veo 3.1.
This is the most important technical feature of Flow AutoKuak for creators producing narrative series. In Sequences mode with a different image per prompt (typical when the flow is Nano Banana → Veo, with the storyboard generated first and the video after), Flow AutoKuak re-selects the correct image on EVERY iteration.
Why it matters
Flow forgets the image-prompt association when you change prompt. If you fire 30 prompts with 30 different images by hand, you have to re-select the correct image 30 times — and one slip costs you 30 videos with the wrong image. Older versions of similar extensions only selected the image once at the start, which meant clips coherent with the first storyboard and random with the rest. Flow AutoKuak fixes this: the image is explicitly re-paired before every send.
Two image modes to accompany your batch depending on the flow (text → video or image → video).
One fixed image (or none) for every prompt in the run. For series sharing the same visual reference on every generation.
Useful for: keeping a consistent look without switching references between clips.
A different image per prompt, assigned in order. Essential when the flow is image → video and every scene has its own storyboard. Flow AutoKuak re-pairs the correct image before every send.
Useful for: numbered storyboards, faceless series with storyboards generated by Nano Banana Pro, long narrative sequences.
Practical rule: if your prompts are short and image-only (Nano Banana), 30 s usually works. If they're long cinematic Veo 3.1 videos, raise the timers to 60–120 s — burning a few extra seconds is worth it to avoid errors that force you to relaunch the batch.
The three automations work the same way (same bridge-panel, same Normal/Sequences modes) and a single Pro plan covers all of them. The difference is the generation engine behind each.
| Grok AutoKuak | Flow AutoKuak | ChatGPT AutoKuak | |
|---|---|---|---|
| Engine | Grok Imagine (xAI) | Veo 3.1 / Nano Banana Pro (Google) | ChatGPT (OpenAI, GPT-4o/4.1) + Sora |
| Output type | Short images and videos | Cinematic video + Nano Banana image | Structured text + Sora image |
| Anon quota | 15 prompts/day | 15 prompts/day | 15 prompts/day |
| Free quota | 30 prompts/day | 30 prompts/day | 30 prompts/day |
| Pro quota | unlimited prompts — each tool tracks its own usage | ||
| Valid pages | grok.com/imagine |
labs.google/fx/.../tools/flow/... |
chatgpt.com |
| Best for | Quick batches, vertical content in bulk | Cinematic production, premium faceless, long narrative sequences | Scripts, SEO, descriptions, chained Sequences |
In practice serious creators use all three: ChatGPT AutoKuak to write scripts and storyboards, Flow AutoKuak to produce the cinematic Veo 3.1 cuts, and Grok AutoKuak for fast experimentation and vertical content. Pro is unlimited on all three and each one tracks usage separately, so alternating doesn't cost the flow anything.
Every tool in the suite tracks its own usage — using Flow doesn't subtract quota from Grok or ChatGPT.
Anonymous
15/day
No account, no sign-in, no card. Enough to try the extension thoroughly.
Free · with Google
30/day
Sign-in via Google OAuth 2.0 + HMAC token between web and extension. Enough for continuous small-scale production.
Pro
∞unlimited
Unlimited prompts on Flow AutoKuak + Grok + ChatGPT. Monthly $9.99 · Quarterly $26.99 · Yearly $89.88 USD. Secure payment via Stripe. Cancel in 2 clicks.
Flow AutoKuak is one of the four extensions in AutoKuak Suite. A single Pro plan covers them all.
What people ask us before installing Flow AutoKuak.
Flow AutoKuak respects your Google Labs / Flow account limits (Veo credits depend on your Google plan). What the extension adds on top is its own usage quota: 15/day anonymous, 30/day Free, unlimited on Pro. If Google caps you at X generations, Flow AutoKuak doesn't multiply them — it automates sending up to your account's limit.
No. Prompts go directly from the browser to Google Flow, and Veo videos download to your local folder via chrome.downloads. The Kuaklabs backend (api.kuaklabs.com) only keeps aggregated usage counters, never the content.
Flow AutoKuak is Manifest V3 and declares only the necessary permissions: webRequest and downloads (to capture and save generated videos) and debugger (to guarantee reliable interactions with Flow's DOM elements). No permission is used outside the Flow domain. Web ↔ extension SSO uses an HMAC token, no plaintext credentials.
We keep selectors up to date as part of the subscription and ship hotfixes via Chrome Web Store within hours/days. To avoid the model-rename problem (Flow renames Veo fairly often), Flow AutoKuak auto-detects the available models on each session instead of hardcoding a list.
Veo 3.1 typically takes 60–120 seconds per short clip. With the recommended 90 s delay between prompts, 30 prompts is ~45 minutes of unattended processing. Split-screen lets you check progress at a glance while you do something else.
Yes — that's the main technical improvement of Flow AutoKuak. In Sequences mode with a different image per prompt, the extension re-selects the correct image on EVERY iteration before sending. If you fire 30 prompts with 30 storyboards, the 30 resulting videos are paired to their correct image, not to the first one.
Yes, it's the most efficient flow. Generate the script and storyboard list in ChatGPT with ChatGPT AutoKuak in Sequences mode (script → scenes → visual prompts). Then capture those visual prompts in Flow AutoKuak (Step 1: source = ChatGPT tab) and fire them at Veo 3.1. Both extensions share the Pro plan and track usage separately.
Yes. Manifest V3, Chrome 116 or newer. Works on modern Chrome, Edge, Brave, and Opera. The extension ships from the Chrome Web Store and behaves the same on all four browsers.