HelloFresh production dashboard
Built around Kelvin and Creative Force for the internal retouching team. The team adopted and used the dashboard.
AI systems architect · Creative technologist · Creative director
I build creative systems, from visual direction to the tools a team uses. My work spans production dashboards, browser workspaces, MCP interfaces and agent workflows.
At HelloFresh, the internal retouching team adopted my Kelvin and Creative Force dashboard. My ANKLE and TITLES work connects creator tools with artist-trained models. Here, voice and speech timing become a character performance you can inspect.
Choose one of five existing voice presets. Press Play to hear the cached recording alongside the 4:5 Kling video loop. New text is synthesized only when you press [SYNTHESIZE].
[LOADING VOICE LIST]
Uses an existing cached result when available; otherwise requests a new ElevenLabs recording. No automatic synthesis or retry. Live character conversation remains paused.
[KLING VIDEO LOOP] Visual accompaniment; no claim of speech-synchronized lip movement.
Voice, timing and a character performance in one workspace. Review the character framing below. Voice is unavailable in this revision; live questions are paused.
“Hi. Hello there! I'm ready for your audio test. What would you like to explore today?”
Saved ElevenLabs audio-test take. It includes a brief interrupted greeting followed by the test response; it is not an answer to a new question.
A selection of the interfaces and visual work behind this application.
Built around Kelvin and Creative Force for the internal retouching team. The team adopted and used the dashboard.
Browser tools that bring visual composition, asset selection and agent workflows into a workspace a creator can inspect and use.
Open workspace [OPEN]Painting, design, image-making and sculptural work. A visual practice that informs the tools and the creative decisions behind them.
Open visual archive [OPEN]MCP interfaces and repeatable workflows for creator studios, including work with artist-trained models at TITLES.
Open developer interface [OPEN]Agent runtime and tool execution work. The link is to the repository; attribution and history remain visible there.
Interactive canvas work connecting reference imagery, graphics and browser controls.
Agent and onchain art workflow experiments. Review the code and repository history.
Eleven v4 Turbo is the speech model. The existing ElevenAgents integration uses a separate reasoning model, PCM audio and provider alignment to drive the character. The ~100ms figure in ElevenLabs’ announcement is a provider inference benchmark, not a measured latency for this application.