With 26,000 GitHub stars and consistent victories over ElevenLabs in blind evaluations, Chatterbox delivers state-of-the-art text-to-speech with zero-shot voice cloning requiring only 5 seconds of reference audio. The model family spans three architectures: Chatterbox Multilingual V3 (500M parameters, 23+ languages including Arabic, Chinese, Japanese, Korean, Hindi, French, German, Spanish, and Portuguese), Chatterbox-Turbo (350M parameters optimized for voice agents with a single-step distilled decoder achieving ~200ms time-to-first-speech), and Chatterbox-Nano (110M parameters running 3x faster than realtime on 8 CPU cores for edge deployment). Unique among open-source TTS systems, Chatterbox introduces emotion exaggeration control — adjusting intensity from monotone to dramatically expressive via a single parameter — and native paralinguistic tagging where tokens like [laugh], [cough], [chuckle], and [gasp] inject natural vocal reactions inline without post-processing. The alignment-informed inference pipeline eliminates hallucinations and repetition artifacts common in autoregressive TTS. Built-in PerTh neural watermarking embeds imperceptible forensic identifiers in generated audio for provenance tracking. Trained on 500,000 hours of cleaned speech data across all supported languages. Voice conversion scripts enable transforming existing audio into any cloned voice. Deploy via pip install with PyTorch, serve through Gradio interfaces or custom FastAPI endpoints, and expose via HTTP streaming or WebSocket for sub-200ms conversational applications. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. MIT licensed.

Kokoro-FastAPI turns text into natural-sounding speech across eight languages by serving the 82-million-parameter Kokoro-82M model through an OpenAI-compatible REST API, so any existing OpenAI SDK client can generate audio by just changing the base URL. With over 5,300 GitHub stars since December 2024, the fully Dockerized FastAPI server covers American English, British English, Spanish, French, Hindi, Italian, Japanese, Brazilian Portuguese, and Mandarin Chinese with language-specific phoneme processing. Inline voice mixing blends multiple profiles using weighted ratios like af_bella(2)+af_heart(1), automatically normalizing weights and caching combined voicepacks as PyTorch tensor files for reuse. Audio streams in real time over HTTP with configurable chunk sizes, or generates complete files in MP3, WAV, OPUS, FLAC, AAC, or PCM formats with speed control from 0.25x to 4.0x. Per-word timestamped captions with speaker-tagged voice labels enable subtitle generation for podcasts, audiobooks, and accessibility workflows. Pre-built Docker images support NVIDIA GPU acceleration via CUDA, experimental AMD GPU inference via ROCm, and CPU-only deployment on linux/amd64 and linux/arm64 architectures, with Apple Silicon MPS support available through direct UV execution. The integrated web interface at port 8880 provides browser-based speech generation, while the Swagger UI at /docs exposes the full API reference. Debug endpoints report system statistics for monitoring inference load. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. Apache 2.0 licensed.

Folivox turns an ebook sitting on your server into a finished audiobook read aloud in a voice you choose, cloned from a short sample, or performed by a whole cast of characters. Drop in an FB2, EPUB, DOCX, PDF, or TXT file and the studio detects the book language, restores chapters, poems, and epigraphs, cleans page furniture away, and spells numbers, dates, money, and abbreviations out in full for nine fully supported languages. Narration runs locally through CosyVoice 3 or the multi-speaker Fish Audio Drama 3, which can give each character their own voice like a radio play, or through ElevenLabs, OpenAI, Gemini, or Cartesia with your own key after a cost estimate you approve. A markup editor finds the dialogue, assigns lines to speakers, and lets you fix mistakes and listen before committing to a ten-hour render. Every fragment is listened back to with Whisper and re-read until the best take survives, while a crash-safe journal means a stopped render resumes where it left off and edited chapters redo only what changed. The finished book arrives as one chaptered M4B with cover and metadata, or one MP3 per chapter, loudness-normalised and ready for any player. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. Apache-2.0 licensed.

OpenTranscribe is a self-hosted transcription platform that converts your audio and video library into searchable, speaker-labeled transcripts, then lets you summarize and chat with those recordings. Upload files up to 15 GB, record from the browser microphone, or paste a link from YouTube and a thousand other sites, and playlists queue up for batch processing automatically. WhisperX produces word-level timestamps in over 100 languages, while the diarization engine works out who spoke when, corrects turn boundaries, and recognizes the same voice again across different recordings. Speaker profiles accumulate a voice fingerprint, speaking-time analytics, and merged identities, so a regular guest is named correctly in every episode without relabeling. Point it at an LLM and recordings gain structured summaries with action items, suggested tags and collections, and a chat interface that answers questions about your transcripts with citations, grouped into projects that pin their source recordings and standing instructions. Leave the model unconfigured and transcription, diarization, full-text and semantic search, and redaction still run entirely on local models. Watch folders, S3 buckets, and network shares import new media on a schedule with three-layer deduplication and automatic stitching of split recordings, and everything exports as subtitles, Word or PDF documents with citeable numbered lines, or Markdown. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. AGPL-3.0 licensed.

ScribeCrate is a self-hosted speech-to-text API that turns recordings into transcripts, subtitles, and English translations on your own server. It speaks the OpenAI Whisper API fluently: point any compatible client at your server's address, swap in your API key, and uploads come back as text. Ask for plain text, JSON with timestamps, or ready-to-publish SRT and WebVTT subtitle files, and for long recordings the segments can stream back over SSE as they are decoded. Turn on speaker diarization and every segment is labeled with who is speaking, which makes interview and meeting transcripts readable at a glance. Choose the model that fits the job, from tiny for quick drafts to large-v3 for maximum accuracy, and run it on a modest CPU, a Raspberry Pi class ARM board, or an NVIDIA GPU. Almost any recording works: MP3, M4A, WAV, WebM, OGG, and FLAC are all accepted, and with models cached in advance the server keeps transcribing with no internet connection at all. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. MIT licensed.

Need subtitles but not a cloud upload? Anysub transcribes audio and video with Whisper on your own server and hands back editable cues. Drop in a video or audio recording, pick a Whisper model from tiny to large, and the worker transcribes it with word-level timing. The studio then puts the video, the cue list, and a zoomable waveform timeline side by side, so every subtitle can be dragged, trimmed, split, merged, or retyped. Warnings flag cues that run too fast, too long, or overlap a neighbor, and autosave with undo keeps edits safe. Long uploads resume after a dropped connection, and the searchable library finds projects by name or by the words spoken inside them. Finished tracks export as SRT, WebVTT, plain text, or JSON with words and metadata. Administration pages manage model downloads, transcription defaults, upload limits, disk usage, and connected workers, and a first-run setup code protects the owner account. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. AGPL-3.0 licensed.

Speakr transforms audio recordings into organized, searchable, AI-enhanced notes with speaker recognition that identifies who said what across your entire recording library. The Python/Flask backend with Vue.js 3 and Tailwind CSS frontend deploys via Docker on port 8899, offering multiple transcription engines through auto-detected connectors: WhisperX for local processing with speaker diarization and voice embeddings, OpenAI Whisper and GPT-4o-transcribe, Mistral Voxtral for cloud diarization, AssemblyAI for multi-hour files, and any custom ASR webservice. Speaker voice profiles use embedding comparison to recognize individuals across different recordings automatically, while custom vocabulary biases the transcriber toward domain-specific jargon. The AI layer goes well beyond transcription: customizable summaries with per-recording, per-tag, and per-folder prompt templates; event extraction surfacing action items and calendar events; per-recording chat with streaming responses; and Inquire Mode for semantic search and natural-language queries across your entire library simultaneously. Smart tags execute custom AI prompts on transcripts for automatic categorization. The REST API with Swagger documentation supports signed webhooks integrating with n8n, Zapier, and Make. Auto-export pushes to Obsidian and Logseq, auto-processing watches directories, and the installable PWA provides mobile-first, offline-capable access with share-target support. 3,600+ stars since May 2025. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. AGPL-3.0 licensed.

With over 48,000 GitHub stars and monthly releases since March 2023, LocalAI is the self-hosted AI engine that replaces every OpenAI endpoint with a single Docker container running on your own infrastructure — serving chat completions, image generation, text-to-speech, speech-to-text, embeddings, vision, video generation, and function calling through identical API schemas that require zero application code changes. The composable backend architecture isolates each inference engine as a separate gRPC service running in its own OCI container, so llama.cpp, vLLM, SGLang, transformers, whisper.cpp, diffusers, MLX, Stable Diffusion, and Flux install on demand without touching the core, can run on separate machines, and a fault in one never affects others. Hardware acceleration spans NVIDIA CUDA 12 and 13, AMD ROCm, Intel oneAPI/SYCL, Apple Silicon Metal, Vulkan, and NVIDIA Jetson L4T — or runs entirely on CPU without any GPU. Built-in AI agents support autonomous tool use, retrieval-augmented generation, Model Context Protocol integration, and skill-based workflows directly in the web interface. The model gallery provides curated YAML configuration files for hundreds of models that install with a single command, while P2P federated inference distributes model shards across multiple machines for running models larger than any single node's memory. Multi-user API key authentication with quotas and role-based access enables team deployments. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. MIT licensed.

OpenVoiceUI is the voice-first front end for a self-hosted AI agent that lives entirely on your own hardware; you talk, it answers out loud, and it builds live pages on screen while the conversation runs. Hold a hands-free conversation with wake word, push-to-talk, and the ability to interrupt, or switch to text when the room is quiet. Speech recognition runs in the browser for free or through Deepgram, Groq Whisper, or your own Whisper server, while nine speech providers from local Supertonic to ElevenLabs give the agent its voice, cloned voices included. As it talks, the agent opens live canvas pages, dashboards, galleries, reports, and tools, and files them on an OS-style desktop you can reopen from the start menu. Ask for music and it generates songs with a built-in player and playlists; ask for images and it creates those too, alongside an audio clip maker and a soundboard. Profiles swap the agent's persona, voice, and model provider, animated faces react while it speaks, and sub-agents and workspace files stay visible in the admin dashboard. The agent itself runs in the OpenClaw gateway with your choice of model provider, so your keys, conversations, and files never leave your server. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. MIT licensed.

Every AI agent gets a face, a colour, and a job in Murmur, a chat app that makes running Hermes Agent as friendly as texting a colleague. Start from a template (Researcher, Writer, Planner, Coder, Coach, or Chef) or describe a new agent from scratch, then talk to one at a time or drop several into a group room and call on a single one with an at-mention. Replies stream in live, long jobs show a plan that ticks itself off, and every step appears in plain words. Connect a model with a ChatGPT plan, a local model server, or an API key from OpenRouter, OpenAI, Anthropic, and others, with the model and thinking effort chosen per agent or chat. Agents can reach Notion, Linear, Todoist, Figma, and dozens of other apps, and they ask for an Allow or Don't allow before changing anything important. Morning briefings run as routines, a memory view shows what each agent knows with controls to edit or forget it, and WhatsApp or Telegram can be linked by scanning a code. A voice call mode lets an agent answer out loud, and slash commands set long-running goals or ask quick side questions. Chats sync across browser, desktop, and phone. Solo builders, homelab fans, and small teams get a calm front door to their own agents. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. MIT licensed.

AudioMuse-AI transforms your self-hosted music collection into an intelligent sonic discovery universe that generates groove-aware playlists without relying on tag metadata or external streaming APIs. The platform scans audio files directly to extract acoustic properties, rhythmic cadence, harmonic density, and lyrical themes across seventy-two supported languages. Listeners can explore their library through a dynamic two-dimensional interactive music map, discover hidden sonic connections between distant genres, or chart smooth transitional song paths between two completely different tracks. Custom prompt-based querying lets users request niche vibes like late-night lo-fi piano, while automated clustering groups acoustically similar tracks into fresh, genre-defying collections. With Song Alchemy, curators fine-tune playlist moods by interactively adding or subtracting acoustic traits before exporting clean tracklists back to Navidrome, Jellyfin, LMS, Lyrion, Emby, or Plex. A built-in acoustic fingerprinting engine even identifies mystery tracks from twenty-second audio snippets, deduplicating songs across multi-server topologies so every file is processed only once. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. AGPL-3.0 licensed.

Every morning, TLDR Radio converts the day's TLDR newsletters into chapterized podcast episodes you listen to instead of read, generated entirely on your own server. Pick a date at the Broadcast Desk, hit Broadcast, and a few minutes later chapterized episodes wait in your browser, one per newsletter with each story as its own track. All fourteen TLDR newsletters can sit on the Desk for manual runs, join the overnight build, or both, and the overnight selection shows an estimated listening time measured from your own episodes. Scheduled builds retry until the day's issue appears and can send an ntfy notification when the audio is ready. The player groups your library by day, shows each chapter with its section heading, source summary, and link to the original, and offers 15-second skips, playback speed up to 2.5x, auto-advance, resume, and keyboard shortcuts. Sponsor segments become their own clearly marked chapters in a different voice, announced out loud so an ad is never mistaken for news, and easy to skip. Sixty-eight voices can be auditioned in Settings, and every episode downloads as a single tagged MP3 with album, title, and author set for podcast and audiobook apps. Running on a dedicated VPS on RepoCloud with guaranteed CPU, RAM, and SSD, full root SSH access, and a browser serial console. MIT licensed.