CrispASR: a unified C++ speech engine
Grown from a whisper.cpp fork: one C++ binary with 119 backends, 62 of them TTS engines.
Verified
Core facts
Read by a script from the official API, not rewritten by a model.
- Repository
- https://github.com/CrispStrobe/CrispASR
- License
- MIT
- Languages
- C++
- Maintainer
- CrispStrobe
- Verified
What it is
CrispASR is one C++ binary with 119 backends, 62 of them TTS engines, plus multilingual text translation and zero Python dependencies. It started as a fork of whisper.cpp. It extends that base into a unified speech engine called crispasr. The engine is backed by full ggml C++ runtimes for major open-weights ASR and TTS architectures.
Main features
You can pick the backend at the command line or let CrispASR auto-detect it from your GGUF file.
It needs no Python, no PyTorch, and no separate per-model binary.
The WebAssembly build is multithreaded and runs entirely client-side with COOP/COEP headers.
Source separation splits a mix into stems.
Pitch estimation gives a monophonic pitch track via CREPE (MIT).
Getting started
The Windows + NVIDIA build is self-contained, and a CUDA Toolkit install is not required.
The macOS build for Apple Silicon uses Metal GPU and Accelerate.
Adding the verbose option to any command shows verbose progress.
Running the list-backends command shows the feature matrix live.
crispasr --version # Windows: .\crispasr.exe --versioncrispasr --backend kokoro -m auto --tts "The quick brown fox jumps over the lazy dog." --tts-output hello.wav
# crispasr: TTS output written to 'hello.wav' (78000 samples @ 24000 Hz, 3.25 sec)crispasr --backend parakeet -m auto -f hello.wav -l en
# crispasr: transcribed 3.2s audio in 0.32s (10.1x realtime)
# The quick brown fox jumps over the lazy dog.Who it is for
CrispASR ships 62 TTS engines for synthesis. Each backend declares capabilities at runtime; if you ask for a feature the selected backend does not support, CrispASR prints a warning and silently ignores the flag. The vulkan builds require the matching driver and do not fall back to CPU.
The Linux cuda tarballs do fall back. Windows CUDA packages bundle the runtime they were built against. The VAD option uses the default Silero VAD (about 885 KB, auto-downloaded).
Turning the LID backend off skips LID entirely. Building CrispASR from source requires the ggml submodule. Selecting the Vulkan backend does not choose which physical GPU to use.
The CrispASR license is MIT, the same as upstream whisper.cpp. Per-model weights are covered by their respective HuggingFace model licenses. CTC-based backends output lowercase without punctuation.
Canary 180M Flash does not provide language detection or native streaming. The server endpoint supports a response format for structured speaker-labelled output. For speaker diarization, vad-turns is a mono-friendly gap-based proxy method.
CrisperWeaver is a cross-platform Flutter transcription app built on CrispASR. Susurrus is a Python ASR GUI with 9 backends.
Related tools
Related tools listed on this site.
百灵大模型
Ant Ling is a language and multimodal model family. Official documentation describes online multi-turn conversations, text generation/content creation, API access and cross-modal understanding/generation.
Exa AI
Exa is a web-retrieval service for AI applications and agents, describing search, content retrieval, deep research and structured-output APIs.
秒听
Tinrec is a meeting-notes and speech-to-text tool. Its public site describes live recordings, audio/video files and video-link transcription, followed by transcripts, meeting notes, speaker summaries and action items.
讯飞译制
iFlyrec Yizhi is a tool for audiovisual translation, AI dubbing and subtitle production, with timecode matching and output described on its official page.
Trancy
A reading and video assistant for language learning. Its site describes bilingual subtitles, selected-text and full-page translation, sentence analysis and personal learning materials.
通义听悟
The site calls it an AI assistant for work and study, built on the Qwen model and audio/video AI, for recording and reading audio/video content. Docs say individuals can use it free after phone sign-up; enterprises can integrate via API.
SoundView
Video-localization tool whose official summary describes translation, dubbing, speech generation, lip sync and subtitle processing.
魔音工坊
Moyin is an AI text-to-speech tool whose official page describes voiceovers for short videos, audiobooks and other narration uses.
讯飞智作
iFLYTEK Zhizuo is an audio/video content-creation platform. Its official site describes text-to-speech, digital-presenter videos and custom appearance/voice assets, including workflows starting from scripts or PPT slides.
逗哥配音
Douge is an AI voiceover tool for short-video production; its official summary also describes script extraction and vocal separation.
ElevenLabs
Text-to-speech and voice selection for narration, character audio, and spoken content.
琅琅配音
Langlang is an AI text-to-speech tool whose official summary describes multilingual voices and emotional styles for uses such as audiobooks and short videos.
Sources
This page is written from the official sources below. Open them to check the original text.
- GitHub READMEhttps://github.com/CrispStrobe/CrispASRFetched
Verified ·
This page is written by AI from the official sources above and checked automatically. This site has not tested it. Found a mistake? Tell us.