My AI Tool Stack
A Closer Look At My Current Setup
It’s not about the tools, but AI capability. But yes, you still want to know what I am using day to day in my work as a marketer. I get that.
A little bit about my work:
My tasks are broad: strategizing, copywriting, evaluating customer feedback, planning a photo shoot, writing scripts and shooting video. I need a versatile tool stack. Hope I can inspire you with the following list:
The go-to all-rounder: Claude Opus 4.6
Writes beautifully, is featured-packed, and I use it for practically everything: from calculating the required amount of loft wallpaper for plastered walls (yes, I use AI at home as well), to writing a solid blog concept, to conducting market research.
What it’s lacking: image and video generation.
The token window was limited before with Opus 4.5 (200K tokens), but now with Opus 4.6 this is 5x the size. It has the same token window as Gemini 3 Pro, 1M tokens.
If you don’t know this yet: Token windows are like working memory for a conversation. 200K tokens ≈ 150,000 words (roughly 300 pages of text). This includes: Your messages, responses, uploaded files. Once full, old messages get dropped to make room for new ones.
When it matters: Very long conversations, multiple large files, or massive documents, Claude Cowork and Projects really benefit from this. For regular everyday tasks, you won’t hit the limit.
An example what Claude Opus 4.6 is capable of:
Input: Asking for a Design System, adding SVG files, a PDF brand book, a clever prompt, asking it to ask questions back.
Output: an HTML-file with a Design System and an MD-text file to reuse in my Claude Code projects (I build my own apps). This makes sure every app is styled in the same way.
Or converting a 6-hour task into a Claude (Cowork) Skill that I can reuse many times, and only takes 2 minutes.
Claude Cowork makes changes directly in folders and files on your computer and I am very impressed on what it can do in regular day to day use.
Image bank superpower: Nano Banana Pro
Creates amazing images and also has consistent output. Miles ahead of ChatGPT’s image generator. I use this daily, literally multiple times. It’s easy to control, even with edits.
I am using this to generate all of the images you see here on my Substack. Also, I use Canva Pro’s 4x of 8x AI Image Upscaler to convert standard sized images to high resolution one. Works like a charm.
Video Generation: Veo 3.1
Slightly more advanced and with more control options than OpenAI’s Sora 2, and more consistent output. The sound is also less tin-sounding. Limitation, of course: a maximum of 12 seconds. Some people don’t realise that AI video is still at a very early stage and compute it scarce, minutes-long videos are not an option (yet).
I use this for B-roll in videos, mostly.
An example of a 18th century sketch I brought to life. I use Google Flow to access Veo 3.1.
Speech / Voice-over Audio: Elevenlabs v3
Very realistic, and easy to control, especially in English. Dutch voices sometimes have a Belgian accent.
I use this if I need a quick voiceover for a video and I don’t have time to add my own voice. Also using it as a voice for AI avatars.
AI Avatar for Long-Format Videos: HeyGen
Almost indistinguishable from a real life speaking actor (especially in English). It does require a learning curve to master HeyGen properly, but after that, you can create 10 minute-long videos. You can even clone yourself, including common hand gestures during speaking.
I have experimented with HeyGen while creating e-learning videos with a cloned human avatar. Results were average, you could see clear differences in tone of voice if you know the person presenting. If not, no problem. Nonetheless, promising technology.
An excerpt from a minute long avatar video (demo) in Heygen. Workflow: Nano Banana Pro to generate the setting + model. Adding voice with Elevenlabs v3 alpha, combining the two in Heygen.
🔗 HeyGen
Music: Suno v5
Did you hear about those Spotify AI-bands that bring out a song every single day and are successful with it? It will probably be generated by Suno. V5 adds more layering than v4 and moving towards ‘truly good music’. In v4, it was still a bit gimmicky.
🔗 Suno
Note: almost all of these tools require the paid plan, but you can often try them for free.
Bonus tips:
Claude Skills (reproducible task blocks; they go much further than custom GPT’s and can also be stacked). Also useful in Claude Code.
Claude Code: build any app that you what. Need I say more? I am currently working on building marketing-related apps. More on that soon!
Zapier: using AI in automations. For example: automatically generating a 80% ready blog post with Claude Opus 4.6, based on an e-mail in G-mail, uploaded as a concept in Wordpress. Or: registering an order in e-commerce, labeling it with AI, and putting it in a Google Spreadsheet.
I use good old ChatGPT sporadically. If so, GPT 5.2 Thinking or Pro for deeper research (marketing plans).
P.S. I’m curious, what are your current go-to AI-tools at the moment? (Share in the comment section)
Very nice stack, Wilbert! And an informative read! I use Chat gpt, ghostbase, substackulous and adobe firefly
Cool stack. Now I'm curious about the handoffs. What happens between the tools is usually where quality diverges. Would love to see you break that part down.