Bror at PlentyLabs gave me free credits to test the platform in August. Nobody paid me for this write-up. I made it because people asked for the workflow after I posted the film on LinkedIn. Every frame and clip in it is AI-generated.
The film is a short B2B piece. Seven characters, fifteen frames, animated clips with dialogue, in a 1980s Scandinavian golf setting by a fjord. The point is simple: everyone has the same tools and the same prompts now, so the person deciding what to make and who it is for is the advantage. I wanted to see whether I could make the whole thing on one canvas instead of juggling an image app, a video app and a voice app.
What PlentyLabs is
An AI creation platform where images, video, voiceover and music live in one subscription and one canvas. It feels like a mix of Figma and Canva. You see the whole project laid out, teammates can join live, there is a chat on the side, and presets handle camera moves and lighting. Every prompt gives you up to four samples to compare.
The six steps
- Concept and script. I wrote the story in plain language first, then cut it into fourteen beats of one or two lines each, because video models hold a scene for a few seconds only.
- Cast sheet. One reference prompt per character, seven in total, with the same anti-gloss skin texture wording in each so nobody looked airbrushed.
- Style block. One paragraph describing the film stock look, the colour palette (green, blue, white) and the 21:9 ratio, pasted at the end of every frame prompt so all fifteen frames read as one film.
- Stills. Fifteen frames generated as images first. This is where I spent most of the iteration, because every clip depends on its frame.
- Motion pass. Each still loaded as the start frame for its clip, with the dialogue line in the prompt. Ten-second clips only for the long lines, five seconds for the rest.
- Assemble. Hard cuts on the script beats, about two seconds per clip, white text overlays with one accent colour, the voiceover running across the cuts, and a light grain pass so the clips feel like one stock. The canvas can trim, sequence and sync to sound, with voiceover and licensed music built in. I did my final cut in CapCut.
The prompts
Character prompt
Portrait of a 15-year-old Norwegian teenage boy, sun-lightened dark blond hair in a loose 1980s side-swept cut slightly grown out over the ears, fair skin with light freckles across the nose and cheekbones, natural summer tan, faint sun flush on the cheeks, clear blue-grey eyes with pale lashes, soft rounded jaw, small gap between front teeth visible when he grins, relaxed friendly expression. Wearing a white knitted short-sleeve polo with a thin navy collar stripe, powder-blue pleated golf trousers with a white canvas belt, white leather golf glove on the left hand, white tube socks, white leather saddle golf shoes. Standing on a golf fairway, relaxed neutral pose, looking at camera, natural soft daylight, eye-level medium shot. [style block]Style block
Style: 1980s analog film photography, Kodak Ektachrome grade, vivid teal-shifted greens, deep blue sky, soft warm sunlight, gentle film grain, slight halation on highlights, shallow depth of field, 35mm cinematic look, nostalgic summer atmosphere. Matte natural skin with visible pores, no glossy skin, no beauty filter. Color discipline: saturated green grass, clear blue sky and water, crisp white and powder-blue wardrobe, no other strong colors in frame. No text, no logos, no modern objects, no phones, no neon colors, no plastic water bottles, no contemporary sportswear branding.Frame prompt
A 15-year-old Norwegian teenage boy with dark blond 1980s hair, freckles, matte natural skin, white knitted polo with navy collar stripe and powder-blue golf trousers, peeking over the lip of a golf green bunker, hands gripping the grass edge, wide curious eyes looking straight into camera, playful expression. Behind him two teenage friends in white and powder-blue golf clothes stand mid-putt, out of focus. Rolling green fairway, dark pine trees, bright blue sky. Low camera angle at grass level, close-up on his face. [style block]Video prompt
He looks into camera, lifts his eyebrows, and says in the bright, casual voice of a 15-year-old Norwegian boy speaking English with a light Scandinavian accent: "Did you know?" Then breaks into a small grin. Light wind moves his hair, the two friends behind make one small slow putting motion. Ambient sound: light wind, distant birds, no music. Maintain the exact 1980s film look, colors, grain, and framing of the input image. Slow, gentle, minimal motion. Face, hair, and clothing remain exactly as in the image, no morphing, no extra limbs. No subtitles, no captions, no text on screen.What I learned about credits
Generate two samples per prompt, not four. Spend most of the budget on the main character's reference sheet, because every later frame depends on it, and accept good-enough for background characters. Generate the hero frames first while the retry budget is full.
What still looks wrong
The voice is the weakest part. I described Erik's voice in each video prompt, so every clip generated its own version and they don't quite match from cut to cut, worst on the frames where you can't see his face, like the top-down cart shot. That was a workflow mistake, not a tool limit. PlentyLabs has Google TTS and ElevenLabs voices built in, and if I did this again I'd pick one ElevenLabs voice sample at the start and use it for every line.
The crowd of friends drifts between frames, a couple of the golf swings have that weightless AI timing, and the transitions rely on the edit in CapCut to hide seams a real camera move would carry. With one voice track and one more refinement pass per shot, this would hold up at full screen.
The workflow as a PDF
The six steps and all the prompts in one file: Download the workflow. Free, no signup.
The film is also on LinkedIn.