Self-hosted generative media on fal.ai — a small, friendly Streamlit app with per-profile libraries, in English and Russian.
git clone https://github.com/evvyraine/vertigo.git
cd vertigo
uv sync
export FAL_KEY="your-key-here" # or add it later in Settings
uv run vertigoRuns on http://localhost:8501. Extra flags are forwarded to streamlit run,
e.g. uv run vertigo --server.port 8501. See
docs/hosting.md for services, Docker, HTTPS and Telegram
sign-in.
- Images — generate with GPT Image 2.5 / Nano Banana 2, edit with references.
- Vectors — Recraft vector art, or trace an image into a real SVG.
- Enhance — upscale (Recraft, Topaz) and remove backgrounds (Ideogram).
- Speech — ElevenLabs text-to-speech and Scribe transcription (with
.srt). - Library — generations run in the background and are saved locally to browse, rename, reuse and delete.
The first run creates three profiles (User 1–User 3, PIN 0000), each with
an isolated library under ~/.vertigo/users/<id>/. Change names and PINs in
Settings; set VERTIGO_HOME to relocate all data. See
docs/hosting.md for a full configuration reference.
Telegram sign-in can replace the PIN gate: each account is mapped to a profile
(created on first sign-in). Set VERTIGO_OIDC_ENABLED=1 with
VERTIGO_OIDC_CLIENT_ID and VERTIGO_OIDC_CLIENT_SECRET from @BotFather, plus
VERTIGO_PUBLIC_URL, and register ${VERTIGO_PUBLIC_URL}/oauth2callback as an
allowed URL in @BotFather.
- Hosting & deployment — services, Docker, reverse proxies, security checklist.
- Contributing
MIT — see LICENSE.