From a YouTuber's self-hosted AI workspace to Alibaba's internal code reviewer, these open-source projects prove that the best software doesn't always come with a price tag.
The open-source movement has never been more potent. In the first instalment of this series, we explored ten remarkable GitHub repositories capable of replacing costly subscription services - and the response was overwhelming. The demand for a sequel was undeniable.
This second volume raises the stakes considerably. The contributors include the world's largest YouTuber, one of the biggest technology conglomerates on the planet, and an e-commerce giant that open-sourced its internal tooling mere days before publication. Every star count has been verified this week, and not a single project overlaps with our previous list.
The financial mathematics are compelling. The combined monthly cost of the proprietary tools these repositories replace exceeds $450 per user. For a small team of five, that approaches £22,000 annually. The open-source alternatives here cost nothing beyond the hardware to run them - and many require nothing more than the laptop you already own.
Let us examine each project in detail, with honest caveats where they exist.
1. Odysseus - PewDiePie's Self-Hosted AI Workspace
Odysseus represents PewDiePie's entry into the developer tools space. Released to immediate fanfare, the repository amassed an extraordinary 20,000 GitHub stars within its first 24 hours - a figure that speaks both to the creator's vast audience and to the tool's genuine utility.
Odysseus functions as a fully self-hosted AI workspace, designed as a direct alternative to ChatGPT Plus and Pro subscriptions. It supports local model execution through vLLM, llama.cpp, and Ollama, whilst also accommodating users who prefer to bring their own API keys. The feature set is ambitious: AI agents with web browsing, file access, and shell commands; deep research capabilities; and automated email triage.
The architecture reflects a growing philosophy amongst privacy-conscious developers: your prompts, documents, and workflows should never leave your machine. For creators handling sensitive material - legal documents, unpublished manuscripts, proprietary business data - this local-first approach isn't merely convenient; it's essential.
Replaces: ChatGPT Plus/Pro (£15–£155/month) GitHub: github.com/pewdiepie-archdaemon/odysseus (opens in a new tab)
2. DeerFlow - ByteDance's Research Engine Without Quotas
ByteDance, the Chinese technology giant behind TikTok, has open-sourced DeerFlow under the permissive MIT licence - and it fundamentally reimagines how automated research should function. Where ChatGPT Pro's deep research feature operates behind strict usage quotas that can leave power users stranded mid-project, DeerFlow offers an agentic harness with no artificial limits.
The technical architecture is genuinely innovative. Each research task executes within its own sandboxed computer environment, ensuring complete isolation and reproducibility. Subagents spawn dynamically on the fly as the research complexity increases, and capabilities are exposed through an intuitive slash-command system reminiscent of Discord or Slack interfaces. Integration channels include both Telegram and Slack, allowing teams to initiate complex research runs directly from their existing communication workflows.
For researchers, journalists, and analysts who routinely hit the ceiling on ChatGPT Pro's deep research allocations, DeerFlow represents not merely a cost saving but a removal of the operational constraints that can stall time-sensitive projects.
Replaces: ChatGPT Pro deep research at full quota (£155/month) GitHub: github.com/bytedance/deer-flow (opens in a new tab)
3. Voicebox - Local Voice Cloning on Your Machine
Voice synthesis technology has advanced rapidly, yet dominant commercial platforms continue charging substantial monthly fees. Voicebox, developed by Jamie Pine, takes a radically different approach - a complete voice studio that runs entirely locally, putting the user in complete control of their audio data.
Voicebox supports seven distinct text-to-speech engines across 23 languages. Voice cloning requires only seconds of sample audio, and the multi-track timeline enables complex productions without leaving the application.
Perhaps most forward-looking is the Model Context Protocol (MCP) integration, allowing AI agents to speak in your cloned voice. This opens fascinating possibilities for automated content pipelines - podcasts, voiceovers, accessibility features - without external API calls or per-character metering.
For podcasters, YouTubers, and accessibility professionals, Voicebox removes both the cost barrier and privacy concerns of cloud-based voice services.
Replaces: ElevenLabs (£17–£75/month) GitHub: github.com/jamiepine/voicebox (opens in a new tab)
4. Open Notebook - NotebookLM Without Google's Limits
Google's NotebookLM has earned a devoted following for its ability to synthesise source materials into coherent discussions, complete with citations. However, the service requires a paid subscription for meaningful usage and imposes structural limitations - notably, its AI-generated podcasts are restricted to exactly two speakers.
Open Notebook replicates and extends this functionality as a fully self-hosted alternative. Users can engage in conversational interactions with their uploaded sources, with every response backed by traceable citations to the original documents. The system supports 18 or more model providers, including local execution through Ollama for those who prefer to keep their documents entirely offline.
The podcast generation feature deserves particular attention. Where Google's implementation fixes the speaker count at two, Open Notebook supports configurations of one to four speakers, enabling more dynamic and varied audio content. A three-person panel discussion or a solo narrated summary becomes possible - structural flexibility that Google's paid offering simply doesn't provide.
For students, researchers, and content creators who process large volumes of source material, Open Notebook delivers the core NotebookLM experience without the subscription cost or the platform lock-in.
Replaces: Google AI Pro (£15.50/month) GitHub: github.com/lfnovo/open-notebook (opens in a new tab)
5. Meetily - Meeting Notes Without the Intrusive Bot
Automated meeting transcription has become standard in corporate environments, yet most solutions introduce an awkward reality: a conspicuous bot joins your call, announcing its presence. Meetily eliminates this friction entirely, running local audio capture and transcription without any visible third-party participant.
With over 234,000 downloads, Meetily has clearly resonated with users seeking a more discreet approach. The transcription engine leverages Whisper and Parakeet for real-time speech-to-text, whilst Ollama handles summarisation - all running locally. Meeting data never traverses external servers unless explicitly configured.
An honest caveat: speaker identification - the ability to automatically label who spoke when - has not yet shipped at the time of writing. The feature is planned for release this month but will be restricted to the Pro tier. Even without it, the core transcription and summarisation represents a compelling free alternative.
For privacy-conscious teams and consultants who cannot risk sensitive client discussions being processed by third-party services, Meetily's local-first architecture is a significant advantage.
Replaces: Otter (£13/month) / Fireflies (£14/user/month) GitHub: github.com/Zackriya-Solutions/meetily (opens in a new tab)
6. Immich - Google Photos on Hardware You Control
Google Photos has long been the default choice for photo backup, yet the service's free tier was effectively eliminated years ago. Users now face a choice: pay a recurring subscription or risk losing their visual memories. Immich offers a third path - a self-hosted photo management platform that rivals Google's offering whilst keeping data firmly under user control.
With 103,000 GitHub stars and a stable v2 release backed by a full-time development team, Immich is no experimental side project. Automatic phone backup over Wi-Fi, machine learning-powered face and object recognition, and plain-English CLIP search all run entirely locally. Search for "dog on the beach at sunset" and Immich surfaces matching images without sending a single byte to external servers.
The project has matured into a genuinely viable alternative for families, photographers, and privacy advocates unwilling to pay perpetual rent or surrender their visual history to a cloud provider's terms.
Replaces: Google One 2TB (£7.50/month) GitHub: github.com/immich-app/immich (opens in a new tab)
7. Dyad - The Lovable Experience, Running Locally
AI-powered code generation tools have transformed software development, but the leading platforms - Lovable, v0 - operate as cloud services with significant monthly fees and inherent lock-in. Dyad replicates the experience of these tools whilst running entirely on your local machine, supporting any model and using your own API keys.
The philosophy is straightforward: your codebase, your choice of model, your control. Dyad doesn't restrict you to a provider's preferred language model or demand ongoing subscription payments for access to your own work.
A transparent caveat applies: Dyad follows an open-core model. The base application is free and fully functional, but a $20/month Pro tier adds credits for enhanced usage and large-codebase modes capable of handling enterprise-scale projects. Even with this premium tier, the cost undercuts Lovable Pro and v0 Team significantly, whilst the free tier alone suffices for many individual developers and small projects.
For developers who appreciate AI-assisted coding but chafe at cloud dependency and recurring fees, Dyad represents a practical middle ground - commercial sustainability for the creators without user exploitation.
Replaces: Lovable Pro (£19/month) / v0 Team (£23/month) GitHub: github.com/dyad-sh/dyad (opens in a new tab)
8. Multica - Managing AI Coding Agents Like Team Members
The emergence of AI coding agents - autonomous systems capable of writing, testing, and refactoring code - has created a new management challenge. How does one coordinate multiple agents effectively? Multica answers this with a management layer that treats AI agents as colleagues rather than tools.
The system's distinctive feature is its squad-based architecture. Users can assign GitHub issues directly to agents, organise them into squads with designated leaders, and oversee operations through a unified interface. The platform supports 11 different agent CLI tools, providing flexibility rather than forcing a single vendor's ecosystem.
The cost comparison is stark. Devin Teams launched at an eye-watering $500 per month before adjusting to $80 base plus $40 per additional seat. Multica delivers comparable coordination at zero licensing cost.
For engineering teams deploying AI coding agents at scale, Multica offers governance and orchestration that would otherwise consume significant budget.
Replaces: Devin Teams (£62/month base + £31/seat) GitHub: github.com/multica-ai/multica (opens in a new tab)
9. Open Code Review - Alibaba's Internal Code Reviewer
In a move that exemplifies the growing corporate commitment to open-source development, Alibaba open-sourced its internal AI code review tool on 10th June 2025 - days before this video's publication. Open Code Review isn't a theoretical project or a simplified subset; it's the actual system Alibaba's engineering teams use internally to maintain code quality across one of the world's largest technology organisations.
The architecture combines deterministic analysis pipelines with an LLM agent, producing line-level pull request comments that identify specific issues rather than vague suggestions. The rule engine covers critical software quality concerns: null pointer exception prevention, thread-safety violations, cross-site scripting (XSS) vulnerabilities, and SQL injection risks.
For development teams currently paying per-developer fees for automated code review, Alibaba's offering represents a credible alternative backed by the engineering rigour of a Fortune Global 500 company. The ruleset reflects real-world production concerns at massive scale - not theoretical academic exercises.
Replaces: CodeRabbit Pro (£19/developer/month) GitHub: github.com/alibaba/open-code-review (opens in a new tab)
10. Dokploy - Deploy to a $5 VPS, Skip the Platform Tax
Modern deployment platforms - Vercel, Netlify, Heroku - have simplified the process of getting applications online, but their pricing models can escalate rapidly as projects grow. Dokploy positions itself as an open-source alternative that runs on commodity virtual private servers, with the tagline "Open Source Alternative to Vercel, Netlify and Heroku."
The capabilities are genuinely broad: support for any programming language, full Docker Compose integration, six built-in database options, and Docker Swarm scaling for horizontal growth. A developer can deploy a full-stack application with database, caching layer, and background workers onto a $5 monthly VPS - a fraction of what managed platforms charge for equivalent resources.
Honest caveats are necessary. Dokploy remains in v0.x versioning, indicating pre-1.0 maturity, and follows an open-core model where advanced features may eventually migrate to paid tiers. Users should evaluate stability requirements carefully before committing production workloads. For side projects, prototypes, and cost-conscious startups, however, the value proposition is undeniable.
Replaces: Vercel Pro (£15/user/month) / Heroku (£5–£1,150/month) GitHub: github.com/Dokploy/dokploy (opens in a new tab)
The Honest Assessment - Caveats and Considerations
Open-source software is not a magic wand that eliminates all costs. Each repository discussed here carries considerations that prospective adopters should weigh carefully.
Time investment is the most significant hidden cost. Commercial SaaS products bundle setup, maintenance, and support into their subscription fees. Self-hosted alternatives transfer these responsibilities to the user. Immich may be free, but configuring backups, managing storage growth, and applying updates requires ongoing attention.
Maturity varies dramatically across this list. Immich, with 103,000 stars and a full-time team, represents enterprise-grade stability. Dokploy, at v0.x, is explicitly experimental. Projects like Odysseus, despite their explosive launch, are young and their long-term maintenance commitments remain unproven.
Support structures differ. When a paid service fails, you file a ticket. When an open-source project breaks, you read documentation, search issue trackers, and potentially fix it yourself. The communities around these projects are generally helpful - but they owe users nothing.
The video's creator, Hyperautomation Labs, deserves credit for surfacing honest caveats for every entry. Speaker identification in Meetily is Pro-only. Dyad and Dokploy are open-core with paid tiers. Some projects are weeks old. This transparency is refreshing in a technology landscape often dominated by uncritical hype.
Conclusion - Choose One, Cancel Something
The ten repositories examined here collectively represent a compelling case for re-evaluating your monthly software subscriptions. From PewDiePie's AI workspace to ByteDance's research engine, from Alibaba's code reviewer to a $5 deployment platform, the breadth of quality open-source software available today is genuinely remarkable.
The practical recommendation is straightforward: select one project from this list that addresses a subscription you currently pay for. Deploy it. Evaluate it honestly over a month. If it meets your needs, cancel the paid alternative and redirect those funds - or simply enjoy the savings.
For those who wish to explore further, Hyperautomation Labs has compiled both instalments of this series into a comprehensive reference document. The open-source ecosystem rewards curiosity and hands-on experimentation. Your next favourite tool may already exist - and it may cost nothing at all.
Helpful Resources
All 10 GitHub Repositories:
- Odysseus (PewDiePie's AI Workspace) - github.com/pewdiepie-archdaemon/odysseus (opens in a new tab)
- DeerFlow (ByteDance Research Engine) - github.com/bytedance/deer-flow (opens in a new tab)
- Voicebox (Local Voice Studio) - github.com/jamiepine/voicebox (opens in a new tab)
- Open Notebook (Self-Hosted NotebookLM) - github.com/lfnovo/open-notebook (opens in a new tab)
- Meetily (Private Meeting Notes) - github.com/Zackriya-Solutions/meetily (opens in a new tab)
- Immich (Self-Hosted Google Photos) - github.com/immich-app/immich (opens in a new tab)
- Dyad (Local AI Code Generation) - github.com/dyad-sh/dyad (opens in a new tab)
- Multica (AI Coding Agent Manager) - github.com/multica-ai/multica (opens in a new tab)
- Open Code Review (Alibaba Code Reviewer) - github.com/alibaba/open-code-review (opens in a new tab)
- Dokploy (Open Deployment Platform) - github.com/Dokploy/dokploy (opens in a new tab)
Free Reference Guide:
- The Free Stack Vol. 2 (PDF with all 20 repos from Parts 1 & 2) - hyperautomationlabs.co/free/free-stack-2 (opens in a new tab)
Paid Tools These Repos Replace (for comparison):
- ChatGPT Plus/Pro - openai.com/chatgpt (opens in a new tab)
- ElevenLabs Voice AI - elevenlabs.io (opens in a new tab)
- Google NotebookLM - notebooklm.google.com (opens in a new tab)
- Otter.ai - otter.ai (opens in a new tab)
- Fireflies.ai - fireflies.ai (opens in a new tab)
- Google One - one.google.com (opens in a new tab)
- Lovable - lovable.dev (opens in a new tab)
- v0 by Vercel - v0.dev (opens in a new tab)
- CodeRabbit - coderabbit.ai (opens in a new tab)
- Devin by Cognition - devin.ai (opens in a new tab)
- Vercel - vercel.com (opens in a new tab)
- Heroku - heroku.com (opens in a new tab)
- Netlify - netlify.com (opens in a new tab)
