A small, edited corner of the internet you can listen to.
Pastel Glass Media is a daily channel that collects long-form writing from Substack and the open web, edits it for the page, and gives every essay a real human voice through neural narration. We curate what's good and we read it out loud.
Why pastel
Most editorial sites today fall into one of two camps. Either they're shouting in primary colors at every reader, or they're so beige and corporate that nothing lands. The internet got loud. We wanted something quieter that still had a point of view.
Each category we cover gets its own pastel accent. It's a small detail, but it makes the site feel like a magazine rather than a feed. Mint for AI, pink for Marketing, lilac for Strategy, sky for Tech, butter for Education, peach for Culture, rose for Fashion, sage for Growth.
The audio
Every essay on Pastel Glass Media ships with full-length narration. Not a 30-second excerpt, the entire piece. We use Microsoft Edge Neural TTS, voiced by Ava Multilingual, and we run it through a chunked pipeline on a dedicated GPU server so the audio cuts cleanly at sentence boundaries.
Why? Because reading isn't the only way to consume an essay. Sometimes you're cooking dinner, on a long walk, or commuting. The web has spent thirty years optimizing for the eye and almost nothing for the ear. We think that's about to change, and we want to be early.
Who curates this
Pastel Glass Media is an editorial side project of Vertical Screen, a small studio across Asia that builds, markets, and publishes. The curation is hand-picked weekly: humble voices, opinionated essays, real arguments. No hot takes, no listicles, no thought-leader fluff.
If you've written something we should republish, or you have a podcast, newsletter, or magazine you'd like to bring under the Pastel Glass roof, write to us.
Built with: Cloudflare Pages, R2, Edge-TTS (Microsoft Neural voices), and a small Python pipeline running on a Forge gaming PC.
Fonts: Archivo Black for display, Source Serif 4 for body, IBM Plex Mono for the masthead, Inter for everything else.
Audio: ~11 minutes average per essay, 48 kbps mono MP3, served from Cloudflare R2.