{"id":51339539,"url":"https://github.com/stremtec/astrape-vst","last_synced_at":"2026-07-02T06:04:12.297Z","repository":{"id":362843822,"uuid":"1261015678","full_name":"stremtec/astrape-vst","owner":"stremtec","description":"Causal zero-shot voice conversion at 44.1kHz.","archived":false,"fork":false,"pushed_at":"2026-06-30T02:50:04.000Z","size":45869,"stargazers_count":24,"open_issues_count":0,"forks_count":1,"subscribers_count":2,"default_branch":"main","last_synced_at":"2026-06-30T03:19:21.676Z","etag":null,"topics":["44khz","causal","neural-audio-codec","q2d2","voice-conversion","zero-shot"],"latest_commit_sha":null,"homepage":"","language":"Python","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":null,"status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/stremtec.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":null,"code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null,"notice":null,"maintainers":null,"copyright":null,"agents":null,"dco":null,"cla":null}},"created_at":"2026-06-06T06:10:25.000Z","updated_at":"2026-06-30T02:50:08.000Z","dependencies_parsed_at":null,"dependency_job_id":null,"html_url":"https://github.com/stremtec/astrape-vst","commit_stats":null,"previous_names":["stremtec/astrape-vc","stremtec/astrape-vst"],"tags_count":0,"template":false,"template_full_name":null,"purl":"pkg:github/stremtec/astrape-vst","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/stremtec%2Fastrape-vst","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/stremtec%2Fastrape-vst/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/stremtec%2Fastrape-vst/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/stremtec%2Fastrape-vst/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/stremtec","download_url":"https://codeload.github.com/stremtec/astrape-vst/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/stremtec%2Fastrape-vst/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":35034989,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-05-26T15:22:16.424Z","status":"online","status_checked_at":"2026-07-02T02:00:06.368Z","response_time":173,"last_error":null,"robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":true,"can_crawl_api":true,"host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["44khz","causal","neural-audio-codec","q2d2","voice-conversion","zero-shot"],"created_at":"2026-07-02T06:04:11.548Z","updated_at":"2026-07-02T06:04:12.290Z","avatar_url":"https://github.com/stremtec.png","language":"Python","funding_links":[],"categories":[],"sub_categories":[],"readme":"# astrape-vst\n\nStrict-causal, zero-shot voice conversion VST — real-time on Apple Silicon (MPS).\n\n| | | \n|---|---|\n| cos768 | **0.935** (probe, 8L StridingAdapter) |\n| Encoder params | 24.9M |\n| Algorithmic latency | ~49ms E2E (0 look-ahead) |\n| Platform | macOS / Apple Silicon (MPS) |\n\n## Design\n\n- **0 look-ahead** — strictly causal. No future frames, KV-cache streaming.\n- **Content/speaker split** — *what* is said (768d content @25Hz) vs *who* says it (global embedding).\n- **Teacher–student** — trained by distilling a frozen, bidirectional MioCodec teacher into a causal student.\n- **Zero-shot VC** — GRL disentanglement + Q2D2 bottleneck. No parallel data needed.\n\n## Quick Start\n\n```bash\n# Voicebank\npython -m astrape.build_voicebank ref.wav -o speaker.astrape\n\n# Convert\npython -m astrape.evaluate --source src.wav --target speaker.astrape --output out.wav\n```\n\n## References\n\n- **Q2D2** — Shuster \u0026 Nachmani, \"Two-Dimensional Quantization for Geometry-Aware Audio Coding\", ICML 2026, arXiv:2512.01537\n- **Mamba** — Gu \u0026 Dao, \"Mamba: Linear-Time Sequence Modeling with Selective State Spaces\", 2023, arXiv:2312.00752\n- **Hyena** — Poli et al., \"Hyena Hierarchy: Towards Larger Convolutional Language Models\", 2023, arXiv:2302.10866\n- **MioCodec** — Aratako/MioCodec-25Hz-44.1kHz-v2 (teacher codec, HuggingFace)\n- **WavLM** — Chen et al., \"WavLM: Large-Scale Self-Supervised Pre-Training for Full Stack Speech Processing\", 2022\n- **GRL** — Ganin \u0026 Lempitsky, \"Unsupervised Domain Adaptation by Backpropagation\", ICML 2015\n- **ConvNeXt** — Liu et al., \"A ConvNet for the 2020s\", CVPR 2022\n- **WavTokenizer** — Ji et al., \"WavTokenizer: an Efficient Acoustic Discrete Codec Tokenizer\", 2024\n- **Snake / BigVGAN** — Lee et al., \"BigVGAN: A Universal Neural Vocoder\", 2023, arXiv:2206.02944\n- **iSTFT / Vocos** — Siuzdak et al., \"Vocos: Closing the Gap Between Time-Domain and Fourier-Based Neural Vocoders\", 2024, arXiv:2306.00819\n- **Predictive Coding** — Oord et al., \"Representation Learning with Contrastive Predictive Coding (CPC)\", 2018\n- **APCodec** — Ai et al., \"APCodec: A Neural Audio Codec\", IEEE/ACM TASLP 2024, arXiv:2402.10533\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fstremtec%2Fastrape-vst","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fstremtec%2Fastrape-vst","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fstremtec%2Fastrape-vst/lists"}