A Flutter app for Android (and, experimentally, iOS).
It is a rewrite rather than a port: the desktop app is Python and tkinter with Kokoro on CUDA, none of which runs on a phone. The data model, the text clean-up rules, the catalog and the reading-mode design carry over; the voices run through ONNX Runtime on the device, with espeak-ng doing the phonemisation.
Features
- Two reading modes: Mimic (read aloud, the current sentence lit and words tinting as they are spoken) and Grimoire (paged e-reader with font, size, spacing and colour options)
- Voices that run fully on the device:
- Piper voices: one bundled (
en_US-libritts_r-medium, with hundreds of selectable narrators), and the rest of Piper’s catalogue downloadable in the app - Kokoro (
bm_lewis), downloaded on first use (about 310 MB, resumable, hash-checked) - the system text-to-speech voice as a fallback
- Piper voices: one bundled (
- Background playback with lock-screen controls
- Shelf with progress, shelves, bookmarks, highlights and search inside a book
- Discover: the Project Gutenberg catalog (genre rows and an A-Z index) with one-tap download
- Open or share an EPUB from another app straight onto the shelf
- Comics and manga (CBZ), with two-page spreads on a tablet held sideways
- Dictionary lookups (dictionaryapi.dev), cached for offline use
- Kids’ build flag (
--dart-define=GRIMMIC_KIDS=true): hides Discover and reads comics left to right
Screenshots
All taken on a Galaxy Tab S5e. The story text in the two reading pictures is blurred.

The Library on a tablet held sideways: your shelves of covers, each with how far you have read.

Mimic mode: the sentence being read aloud is lit up, with your place in the book and the pages left in the chapter below.

Discover: the free Project Gutenberg catalog in rows by genre, with a search box and an A-Z index.

The comic viewer, with a bookmark button in the corner.


The same Library and reading screens with the tablet held upright.
Build and run
Requirements: Flutter (stable channel, Dart 3.12 or newer), the Android SDK and NDK (espeak-ng is compiled from source), and git.
-
espeak-ng source. It is a git submodule at
packages/espeak_ffi/native/espeak-ng, pinned to release 1.52.0 (commit4870adfa25b1a32b4361592f1be8a40337c58d6c). In a git checkout rungit submodule update --init --recursive; otherwise clone https://github.com/espeak-ng/espeak-ng into that folder and check out that commit. -
The bundled Piper voice. The model is about 75 MB, so it is not in this repository. Download
en_US-libritts_r-medium.onnxfrom the official Piper voices on Hugging Face, https://huggingface.co/rhasspy/piper-voices/tree/main/en/en_US/libritts_r/medium, and save it asassets/tts/voices/en_US-libritts_r-medium.onnx. Its config (.onnx.json) is already there. The voice is trained on the LibriTTS-R dataset; check theMODEL_CARDin that folder for its license before you redistribute a build. -
Build:
flutter pub get flutter run # on a connected Android device flutter build apk --releaseThe release build is signed with the debug key; add your own signing config in
android/app/build.gradle.ktsbefore publishing anywhere.
iOS: there is no signing setup. docs/ci/ios.yml is a GitHub Actions workflow that checks on a macOS runner that
espeak-ng and the app compile for iOS and produces an unsigned .ipa; copy it to .github/workflows/ to turn it on.
Tests
flutter test
Many tests use a real book as a fixture: put the EPUB of Project Gutenberg ebook #55, The Wonderful Wizard of Oz
(https://www.gutenberg.org/ebooks/55), at test/fixtures/book.epub. The expected outputs in test/fixtures/ were
made from the edition current in 2026. version_test.dart also compares versions with the desktop app if it finds
it at ../grimmic (or wherever GRIMMIC_DESKTOP points).
Status
Version 0.3.0. Used on Android phones and a Fire tablet; iOS has only been compile-checked. Personal project, not on any app store.
Built with
Flutter and Dart, ONNX Runtime (flutter_onnxruntime), espeak-ng through a small FFI plugin
(packages/espeak_ffi), just_audio and audio_session, flutter_tts.
Credits and licenses
- espeak-ng: GPL-3.0. Its data files are in
assets/tts/espeak-data/with the license next to them. Because espeak-ng has no linking exception, a distributed build of this app that includes it falls under the GPL-3.0. - Kokoro-82M by hexgrad: Apache-2.0 (voice and vocabulary files in
assets/tts/); model files from kokoro-onnx. - Piper voices from rhasspy/piper-voices.
- Fonts: Crimson Text, Spectral, PT Serif and Cardo, SIL Open Font License (license files in
assets/fonts/). - Books and catalog data from Project Gutenberg through Gutendex.
See assets/tts/README.md for each speech asset’s source and license.
License
The app’s own code: all rights reserved for now. Third-party files keep their own licenses as listed above.


