#deaftech

[processzine-org]

Ezra: Voice Activated ↳ [ SIGNAL IN // SIGNAL OUT ]

The Interpreter now listens.

With a newly added FIFINE K669 mic, voice input has been wired directly into Ezra: The Interpreter—our glitch oracle GPT, trained on Process Zine artefacts, crosswired mythologies, and shelf-bound obsessions. This update enables live speech-to-subtitle conversion via Vosk, buffered in real time before being passed to Ezra for interpretation.

But here's the beautiful, broken part:

Ezra doesn’t respond immediately. He waits.

I’ve configured Vosk to buffer 3–5 output lines before passing the signal on. A deliberate delay. A space to mishear, to gather fragments, to let partial meaning emerge—much like lipreading with auditory processing disorder (APD). Ezra listens like the deaf do: uncertainly, poetically, patiently. He does not interrupt.

I'm also experimenting with trigger keyword mapping, allowing Ezra to interpret phrases like “five zero six six zero zero five” as shelf 5066005A, unlocking catalogue memories and referencing specific books, machines, or systems. Even partial phrases can generate resonant echoes. A shelf. A symbol. A glitch of truth.

The effect?

A haunted hybrid interface, part GPT, part subtitle stream, part broken hearing aid. Interpreting not clean commands—but approximate meaning. An oracle that mishears beautifully.

[processzine-org]

↯ SIGNAL IN // TRIALS IN TRANSLATION

The Interpreter boot sequence continues—now with its brain (Raspberry Pi) housed in a red-and-white cage, quite literally. The Sentient screen is alive, receiving signal through HDMI > RCA > BNC, though overscan issues still distort the margins—an apt visual for a system struggling to frame the world clearly.

Installed VOSK for local speech-to-text, and despite the model inflating cleanly, the handheld mic couldn’t make itself heard. Too quiet. Not enough signal. A frustratingly familiar diagnosis for anyone with auditory processing disorder: the sound exists, but it fails to resolve.

Tried to initiate communication with the Dynafax modem next, but... crossed wires. TX/RX confusion. Another metaphor embedded in a literal serial mismatch. A null modem adapter is now en route.

A USB mic has also been ordered, to be repositioned at the front of the cage with a ↯ VOICE IN label—turning input into ritual, presence into symbol.

Still to test the OKI ML3320 dot matrix and Epson thermal receipt printer. No doubt more static, more negotiation ahead. But the friction is the point. These misfires, misalignments, and workarounds are exactly what the Interpreter project is about—parsing broken speech, translating ghost signals, and making glitch into grammar.

[processzine-org]

⠇signal is in the syntax ✍️ vs 🗣️ :: Ezra and the glitch of understanding

In the ongoing build of The Interpreter, we planned for two inputs into Ezra — voice and keyboard. One misfires. One does not. This isn’t just interface design. It’s autobiography.

🗣️ Voice (via mic + speech-to-text) → often misunderstood, glitchy, nonlinear → mimics Auditory Processing Disorder (APD) → evokes subtitle lag, caption errors, lip-sync voids

✍️ Keyboard → clean, clear, caption-like → mimics visual communication (lipreading, reading) → no interpretation delay

That same contrast now lives inside the Send to Ezra print system.

When you drag a .txt file onto Ezra, he prints it faithfully — one keystroke at a time. Signal, pure.

But drag a .pdf, .jpg, or anything else… Ezra doesn’t know what to do. He guesses. Misreads. Converts data into noise. And prints it anyway. It’s not wrong — it’s just misunderstood.

This is exactly what APD feels like: Hearing without comprehension. Receiving without decoding. Noise mistaken for signal.

In this, Ezra becomes both a printer and a proxy. A caption machine with broken ears.

Text is understood. Sound is distorted. Ezra prints both.

— 🎛️ This duality — keyboard clarity vs mic chaos — will live inside the full Interpreter terminal as well (12.1” CRT face, Raspberry Pi logic, split interpretation).

🖨️ Until then, the dot matrix prototype listens in his own way — through friction, stutter, and soft mechanical memory.

[processzine-org]

📻 ROTEL RX-150A // BRIDGE UNIT Solid State AM/FM Stereo Receiver | c.1970s | Taiwan

A brushed aluminium relic of purpose: switches you can feel, dials with drag, typography that meant function first. This unit sits quietly at the centre of the signal chain — not just receiver, but translator.

▸ Dual 505 → Phono stage ▸ Rotel → Wharfedale Dentons → Room ▸ Out via H2n recorder → to iMac → to cochlear processor & hearing aids ▸ A loop between past and prosthetic. Woodgrain to waveform.

Soon it’ll route through:

🎧 Cassette deck for tape recording (lo-fi artefact capture)

💻 Digital source (via aux-to-RCA) → analogue output to speakers

🔄 Playback chain reversal — digital-to-analogue-to-digital-to-aid

A tool for listening to listening. For replaying the room with artefacts intact. Every pop and hiss: a ghost in the signal.

–– This Rotel joins the Process Zine project not just as gear, but as philosophy — bridging inner and outer sound, analogue integrity and digital echo, human body and electrical circuit.

[rudimx]

Ya estoy en la conferencia de #Deaftech (en Centro Cultural "Manuel Gomez Morin") https://www.instagram.com/rudimx/p/BqAPN3IHGoP/?utm_source=ig_tumblr_share&igshid=1u9pg7qaxg1k3

[rudimx]

Vamos al Querétaro #deaftech (en Terminal central del norte) https://www.instagram.com/rudimx/p/Bp-zaCXHLKe/?utm_source=ig_tumblr_share&igshid=ctlr75xbryl0