Songbird

Change the voice in your head

Songbird's premise is simple: the voice guiding a meditation, affirmation, or hypnosis-style script matters as much as the words. So instead of a generic narrator, Songbird generates the audio in a voice the listener actually trusts — their own cloned voice, or one they've chosen and trained specifically for this purpose.

It's built for individuals working on their own mindset and habits, and for coaches and hypnotherapists guiding clients through that work. It's explicitly framed as guided transformation work, not therapy or medical care.

Writing and generating a track

The core editor is a script-writing surface: you write the track's language directly, dropping in bracketed delivery cues — [hopeful], [gentle], [confident], [grounded] — that shape tone as the line is spoken. From there you pick a voice from Songbird's library or a trained custom voice, choose background audio, adjust delivery speed, and queue the track for generation.

Training a voice

Custom voices are trained through ElevenLabs' voice-cloning models on top of user-submitted audio samples. For coaches working with clients, Songbird includes a voice-recording request flow: a coach sends a client a public link, the client records and uploads a few voice clips from their own device, and the coach reviews the samples before the voice is trained and made available for that client's tracks.

Built for individuals and for coaches

Songbird shipped as two distinct plans: one for individuals creating tracks for themselves, and one for coaches and hypnotherapists managing multiple clients — each with their own voice profiles, track libraries, and playlists. The coach view adds a client roster and per-client history on top of the same creation tools individuals use.

The Gnome Grove 🍄

Alongside the standard flow, Songbird has a free, whimsical alternate mode called the Gnome Grove. Instead of a straightforward affirmation form, a small forest guide named Grimblewood Thornwhisker asks what you're carrying on the path today and turns it into a personalized "forest blessing" — read aloud in his own trained voice over a bed of mossy, lantern-lit forest ambience.

It's meant to be Songbird's free tier — a low-pressure way to try the product before ever paying for it — so it runs on a gentle monthly cap. Don't be surprised if Grimblewood is occasionally all out of blessings for the month and asks you to check back later.

A live teaser of the Gnome Grove is coming soon — check back here for a link.

Under the hood

The backend is Go (chi) with Temporal orchestrating the longer-running voice-training and track-generation workflows, PostgreSQL for data, Auth0 for authentication, and S3/MinIO for audio storage. The frontend is React and Vite with shadcn/ui. I built and shipped the whole stack solo — product design, backend, frontend, and the voice pipeline.

Songbird's public site is offline right now while I figure out its next chapter — the recordings above are from the working product.