Image and video with direction.
Visual production with generative tools, but with editorial judgement and human direction. No random prompts, no overheated templates, coherence sustained piece by piece across brands, IPs and narrative universes.
The generative audiovisual production I do doesn't compete with a film shoot or a traditional animation studio, nor does it try to. It competes with stock libraries, Canva templates, fast productions that need brand quality but don't have the budget or time for a big team. That's where, well-directed, generative tools deliver results that weren't possible at this cost three years ago.
What sets my work apart from "someone who knows how to push a button in Midjourney" is three things. One, real brand direction: I work palette, composition, and the project's visual language before generating anything. Two, consistency between pieces: if your character has a certain face in piece 1, it's still the same face in piece 47, extremely difficult, requires LoRAs, character locking techniques, and a lot of curation. Three, real post-production: model output is never the final delivery, there's grading, retouching, editing, sound design.
I have my own projects where I practise this: NIXIA (AI-assisted music with coherent visual identity), Vex Muse, Valentina Serrano, Morgan Serafi (digital personas in continuous development). They're not isolated experiments: they're sustained discipline over months to keep a visual identity that genuinely looks the same in every piece.
How we work
-
01
Brief and art direction
Before generating, we define: which project, which references, which palette, which tone. If your brand already has identity, we start there. If not, we build it first.
-
02
Project-appropriate technical pipeline
Decide which models (Flux, Wan, Veo, Seedance), which pipeline (local ComfyUI, cloud API), which LoRAs to train. The choice depends on target style and volume.
-
03
Iterative production with curation
Generation + rigorous selection. The real ratio is ~1:50, for one good piece I discard forty-nine. That's invisible to the client but it's what separates acceptable from excellent.
-
04
Post-production and delivery
Grading, editing, audio, formats. Delivery in the final formats you need (vertical for social, horizontal for web, master for print).
What's included
- Initial brief and art direction
- Pipeline technical definition (which models, which tools)
- Generation + curation of image, video and audio per project
- Custom LoRA training when the project calls for it
- Post-production: grading, editing, sound design
- Delivery in all required formats
- Pipeline documentation for future reproducibility
When we're not the right fit
- You need a real shoot If your project requires real actors, physical sets or classical cinematography, I'm not a production company, find a traditional audiovisual team.
- "I want the video in 24h" Direction and curation take time. If that's your timeline, better stock or a junior freelancer with loose prompts. I won't compete with that.
- You want a content factory of hundreds of monthly pieces My process is sustained quality, not industrial volume. If you need to feed 200 monthly publications, I can help you set up the pipeline but I won't produce all of it myself.
Stack & tools
- Image models: Flux, Stable Diffusion, Qwen Image, Nano Banana, Seedream
- Video models: Wan, Veo, Kling, Sora, Seedance
- Audio models: ElevenLabs, Suno, Udio
- Local pipelines: ComfyUI + SwarmUI on RTX 4070 Super
- LoRA training for character consistency
- Post: DaVinci Resolve, Audition, traditional editing
- Cloud: RunPod, Replicate when more GPU is needed
If you have a project where you want coherent, sustained visual quality, write to me with the idea, first conversation is free and it usually becomes clear quickly whether we fit.
Let's talk →