Undocumented: TextFace

Over the last months, alongside developing Lessons From Departure: Shattered World, I have been building an experimental visual tool called Text Face.
At a technical level, Text Face is a browser-based system that transforms live video, uploaded footage, webcam feeds, screen capture, or images into a shifting low-resolution wall of coloured text. However at a human level, it is an attempt to answer a more difficult question:

How do we allow people to tell deeply personal stories without demanding full visibility in return?

That question sits at the centre of much of my recent work around migration, displacement, identity, rebuilding, and the fragile space between being recognised and needing protection.

Faces Made From Words

Text Face uses face tracking, local transcription, and text rendering to convert human faces into layered structures built entirely from language.

Instead of seeing a conventional portrait, viewers encounter something more unstable.

A face appears.
Then dissolves.
Words replace skin.
Speech becomes surface.
Identity becomes readable, but not fully visible.
The person remains present, but difficult to consume through immediate visual recognition.

This is intentional.

The tool emerged from a recurring challenge inside collaborative storytelling work:

Many people carry powerful stories they are willing to share, but not everyone can safely appear in front of a camera. For participants navigating uncertain legal, social, political, or personal realities, visibility itself can carry risk.

Traditional documentary forms often depend on showing faces clearly in order to generate emotional connection. I wanted to explore another possibility. What happens if language becomes portraiture?

The Technical System:

Text Face is built as a local-first browser tool.

It currently supports:
Visual inputs
uploaded video
uploaded images
webcam
screen capture
Text inputs
manual text
uploaded transcript assets
live speech

These inputs can be mixed independently.

For example:
webcam + live speech
recorded interview + transcript
face video + manually written testimony

The system uses:
MediaPipe Face Landmarker for face tracking
faster-whisper for local transcription
browser canvas rendering
local Python processing utilities

Transcripts can be generated locally with timestamps and synchronised back into the visual system.

Words can then move through the face itself, highlighting in sync with playback. Importantly, the workflow is local-first. Transcription and processing happen on the operating machine rather than through cloud services wherever possible. For projects involving vulnerable participants, this matters.

Why I Built It

Although Text Face began as a technical experiment, it is increasingly becoming part of a broader artistic investigation.

My ongoing project, Lessons From Departure, works through long-form conversations with people who have experienced migration, displacement, separation, uncertainty, and rebuilding. The portraits in that work intentionally resist easy recognition.

Faces fragment.
Layer.
Disappear.
This is partly aesthetic, but also ethical.
The work asks:

Can someone remain visible without becoming exposed?

Recently, after attending storytelling presentations developed through IHRE / Foundation Academy Amsterdam, another possible direction began to emerge. Participants shared migration experiences through ordinary objects:

a towel
a phone
certificates
a jacket
a plastic bag
a door

The stories explored dignity, survival, memory, adaptation, and life lived between systems. Many participants were speaking from spaces of uncertainty, invisibility, or unresolved status. The overlap with my own work felt immediate.

TextFace suddenly began to feel less like a visual experiment and more like a possible collaborative framework.

Not necessarily to replace documentary portraiture. Not necessarily to hide everyone. But to offer another option.

A way for someone to say:

I want my words present. I want my story heard. I do not necessarily want my face recognised.

Where This Could Go

I am currently exploring how Text Face might support future collaborative storytelling work connected to migration, identity, and human rights.

One possible direction would combine:

participant-led testimony
object-based storytelling
optional visual anonymity
synthetic or original voice choices
projected text portraits
installation-based presentation

I can imagine words projected onto towels, certificates, doors, fabric, printed human-rights texts, or embedded screens. Stories becoming environments rather than simple interviews. The goal is not technological novelty.

The goal is to create more flexible ways for people to participate on their own terms. To allow complexity without demanding exposure. To explore what portraiture becomes when language, memory, identity, and protection occupy the same space.

Text Face is still evolving.

Like much of the work around it, it remains experimental, collaborative, and unfinished.

Which feels appropriate.

Because the stories that inspired it are unfinished too.