Debjit B. (@debjit012) header
Debjit B. (@debjit012) profile picture

Full-stack developer. Passionate about Laravel, React, React-Native, and TailwindCSS. Learning marketing by building in public: https://teapawt.org/t.co/UdehhirHwa

WB, India
Joined September 2012
Photo by Debjit B. (@debjit012)Photo by Debjit B. (@debjit012)Photo by Debjit B. (@debjit012)Photo by Debjit B. (@debjit012)Photo by Debjit B. (@debjit012)Photo by Debjit B. (@debjit012)Photo by Debjit B. (@debjit012)Photo by Debjit B. (@debjit012)Photo by Debjit B. (@debjit012)Photo by Debjit B. (@debjit012)
Paying for a wrapper? Instead, use lipi.debjit.in, OSS. Most of the ASR models are free and OSS; get the best of the free models with my lipi.debjit.in.
I audited 30+ voice dictation apps to see how they actually work—and how they’re priced—while building my own open-source, self-hosted alternative in public. Most AI dictation tools today charge $8–$30/month to wrap speech-to-text models and an LLM cleanup step. As a builder, I wanted to see what happens if you flip that model: make it local-first, support self-hosted models & endpoints, keep it open-source, and learn marketing by shipping in public. That experiment is Lipi (lipi.debjit.in) — a lightweight desktop dictation app for Windows & Linux (Alt+R anywhere to speak and paste directly into your active window) with local inference and self-hosted AI support. Before locking in my roadmap, I mapped the entire dictation landscape by pricing and openness. Bookmark this if you're choosing a tool (or building in this space): 1. Free & Open-Source (Self-Hosted / Local-First) • Lipi (lipi.debjit.in) — Free & OSS (AGPL-3.0). Local-first desktop engine + localized/self-hosted AI cleanup for Windows & Linux. • Handy — Free & OSS (MIT). Local Whisper/Parakeet/Moonshine engines on macOS, Windows & Linux. • OpenWhispr — Free & OSS (MIT). Local on-device models or BYOK cloud APIs across desktop & iOS. • nerd-dictation — Free & OSS (GPL-3.0). Minimalist offline Vosk-based dictation for Linux. • Talon Voice — Free public build (OSS user scripts; $15/mo+ Patreon for early beta). Hands-free voice coding & OS control. 2. Built-In OS & Free Proprietary Tools • Apple Dictation — Free ($0). Built into macOS, iOS & iPadOS (runs on-device on Apple Silicon). • Windows Voice Access / Voice Typing (Win+H) — Free ($0). Built into Windows 11. • Google Docs Voice Typing — Free ($0). Built into Chrome. • Gboard Voice Typing — Free ($0). On-device on Pixel 6+ and cloud on Android/iOS. • Google Recorder — Free ($0). Offline transcription & speaker labels on Pixel. 3. Open-Source / Source-Available (Paid Prebuilt Binaries) • VoiceInk — OSS (GPL-3.0, free if self-built) or $25–$49 one-time for prebuilt Mac app. • Sayboard — OSS / Freemium (~$20–$25 one-time lifetime unlock). • Amical — OSS core / Freemium (unlimited local + 8k cloud words/mo free; $7–$10/mo Pro).• FUTO Voice Input — Source-available / $9.99 one-time on Android (100% offline Whisper). 4. Freemium SaaS & Hybrid AI Dictation • Spokenly — Free local tier; $29 one-time (Local+) or $9.99–$12/mo (Pro). • Superwhisper — Free unlimited local; $8.49/mo, $84.99/yr, or $249.99 lifetime. • MacWhisper — Free tier; €64 one-time Pro license (macOS). • WhisperTyping — Free tier; $5/mo (Personal), $15/mo (Pro), or $30/mo (Medical) billed annually. • BlabbyAI — Free (60 credits/wk); $8.49/mo unlimited. • Typeless — Free (8,000 words/wk); $12–$30/mo Pro. • Wispr Flow — Free (2,000 words/wk); $12–$15/mo Pro. • Willow Voice — Free (2,000 words/wk); $12–$15/mo Individual. • Aqua Voice — Free (1,000 words); ~$8–$10/mo Pro.• Voibe — Free (2,000 words/wk); $99/yr or $199 lifetime. • Voicenotes — Free (100 mins/wk); $9–$14.99/mo Pro. • Letterly — Free (10 notes); $12.90/mo Pro. • Speechnotes — Free (ad-supported); ~$1.90–$9.90/mo or one-time unlock. 5. Paid-Only & Enterprise • Voicy — $8.49/mo, $82/yr, or $260 lifetime (30-min trial). • Monologue — $15/mo or $144/yr (1,000-word trial). • BetterDictation — $39–$149 one-time Mac license (+ optional $2/mo add-on). • Dragon Professional v16 — ~$699.99 one-time (Windows) or ~$55/user/mo (Anywhere). 3 things I learned from this pricing teardown: Almost every $12–$15/mo app is selling convenience + cloud LLM text formatting on top of the exact same open-weight speech models you can run locally. Linux and self-hosted setups are still treated as second-class citizens—even though developers and privacy-conscious users dictate all day. Giving people full local & self-hosted control (running your own local speech models + your own local/self-hosted LLM endpoint) removes subscription fatigue completely. This whole project is an experiment for me—learning product positioning, open-source distribution, and marketing from scratch while building something I use every day. If you're on Windows or Linux and want fast, local, self-hosted dictation without a monthly subscription, try Lipi or star/fork the repo:lipi.debjit.in What’s your current dictation setup—local models, cloud AI, or built-in OS tools?
18
Debjit B. retweeted
Lipi 0.1.8 is live with Meetings (beta). Record your mic + system audio on any call. When you stop, you get a transcript, summary, decisions, and action items saved locally on your PC. Tell everyone before you record. ★ github.com/debjit/lipi #buildinpublic
1
1
1
49
👍
AWS just released an open-source AWS Emulator. The interesting part? 👀 Thousands of AI agents wrote 100% of its code. No human-written or human-reviewed code, according to AWS. And it's free to run locally, without an AWS account.
17
I audited 30+ voice dictation apps to see how they actually work—and how they’re priced—while building my own open-source, self-hosted alternative in public. Most AI dictation tools today charge $8–$30/month to wrap speech-to-text models and an LLM cleanup step. As a builder, I wanted to see what happens if you flip that model: make it local-first, support self-hosted models & endpoints, keep it open-source, and learn marketing by shipping in public. That experiment is Lipi (lipi.debjit.in) — a lightweight desktop dictation app for Windows & Linux (Alt+R anywhere to speak and paste directly into your active window) with local inference and self-hosted AI support. Before locking in my roadmap, I mapped the entire dictation landscape by pricing and openness. Bookmark this if you're choosing a tool (or building in this space): 1. Free & Open-Source (Self-Hosted / Local-First) • Lipi (lipi.debjit.in) — Free & OSS (AGPL-3.0). Local-first desktop engine + localized/self-hosted AI cleanup for Windows & Linux. • Handy — Free & OSS (MIT). Local Whisper/Parakeet/Moonshine engines on macOS, Windows & Linux. • OpenWhispr — Free & OSS (MIT). Local on-device models or BYOK cloud APIs across desktop & iOS. • nerd-dictation — Free & OSS (GPL-3.0). Minimalist offline Vosk-based dictation for Linux. • Talon Voice — Free public build (OSS user scripts; $15/mo+ Patreon for early beta). Hands-free voice coding & OS control. 2. Built-In OS & Free Proprietary Tools • Apple Dictation — Free ($0). Built into macOS, iOS & iPadOS (runs on-device on Apple Silicon). • Windows Voice Access / Voice Typing (Win+H) — Free ($0). Built into Windows 11. • Google Docs Voice Typing — Free ($0). Built into Chrome. • Gboard Voice Typing — Free ($0). On-device on Pixel 6+ and cloud on Android/iOS. • Google Recorder — Free ($0). Offline transcription & speaker labels on Pixel. 3. Open-Source / Source-Available (Paid Prebuilt Binaries) • VoiceInk — OSS (GPL-3.0, free if self-built) or $25–$49 one-time for prebuilt Mac app. • Sayboard — OSS / Freemium (~$20–$25 one-time lifetime unlock). • Amical — OSS core / Freemium (unlimited local + 8k cloud words/mo free; $7–$10/mo Pro).• FUTO Voice Input — Source-available / $9.99 one-time on Android (100% offline Whisper). 4. Freemium SaaS & Hybrid AI Dictation • Spokenly — Free local tier; $29 one-time (Local+) or $9.99–$12/mo (Pro). • Superwhisper — Free unlimited local; $8.49/mo, $84.99/yr, or $249.99 lifetime. • MacWhisper — Free tier; €64 one-time Pro license (macOS). • WhisperTyping — Free tier; $5/mo (Personal), $15/mo (Pro), or $30/mo (Medical) billed annually. • BlabbyAI — Free (60 credits/wk); $8.49/mo unlimited. • Typeless — Free (8,000 words/wk); $12–$30/mo Pro. • Wispr Flow — Free (2,000 words/wk); $12–$15/mo Pro. • Willow Voice — Free (2,000 words/wk); $12–$15/mo Individual. • Aqua Voice — Free (1,000 words); ~$8–$10/mo Pro.• Voibe — Free (2,000 words/wk); $99/yr or $199 lifetime. • Voicenotes — Free (100 mins/wk); $9–$14.99/mo Pro. • Letterly — Free (10 notes); $12.90/mo Pro. • Speechnotes — Free (ad-supported); ~$1.90–$9.90/mo or one-time unlock. 5. Paid-Only & Enterprise • Voicy — $8.49/mo, $82/yr, or $260 lifetime (30-min trial). • Monologue — $15/mo or $144/yr (1,000-word trial). • BetterDictation — $39–$149 one-time Mac license (+ optional $2/mo add-on). • Dragon Professional v16 — ~$699.99 one-time (Windows) or ~$55/user/mo (Anywhere). 3 things I learned from this pricing teardown: Almost every $12–$15/mo app is selling convenience + cloud LLM text formatting on top of the exact same open-weight speech models you can run locally. Linux and self-hosted setups are still treated as second-class citizens—even though developers and privacy-conscious users dictate all day. Giving people full local & self-hosted control (running your own local speech models + your own local/self-hosted LLM endpoint) removes subscription fatigue completely. This whole project is an experiment for me—learning product positioning, open-source distribution, and marketing from scratch while building something I use every day. If you're on Windows or Linux and want fast, local, self-hosted dictation without a monthly subscription, try Lipi or star/fork the repo:lipi.debjit.in What’s your current dictation setup—local models, cloud AI, or built-in OS tools?
16
1
2
14
438
Speak, don’t type. Lipi is a lightweight, local-first open-source dictation tool with Alt+R system-wide pasting into any app. Works offline, keeps everything on your device, and is perfect for accessibility. Download free: lipi.debjit.in #OpenSource #LocalFirst #Accessibility #SpeechToText
1
1
28
Selling daily oral herbal sprays to build immunity, to improve gut health, or to cure allergic rhinitis is completely illogical, scientifically baseless, and potentially dangerous for children below five years of age. This is what one Dr Nihar Parekh is doing in the name of Primerry oral herbal sprays. These are being endorsed by @NehaDhupia and @sakpataudi When a doctor uses their credentials to push these unproven products, they are trading clinical science for commercial success. Here is why this trend is a massive red flag: The human body does not build systemic immunity or realign a complex gut microbiome via a daily micro-mist of herbs sprayed into the mouth. It is a marketing gimmick wrapped in medical jargon. In children under the age of 5, shooting a sudden blast of liquid spray into the back of the throat can rarely trigger laryngospasm —a terrifying protective reflex where the vocal cords go into a spasm, temporarily blocking the child's airway. Young children have highly sensitive gag reflexes. Forcing a daily herbal spray into their mouths is a recipe for choking, vomiting, and severe respiratory distress. Dealing with a child suffering from chronic allergic rhinitis or constant daycare sickness is exhausting. Exploiting this exhaustion, and selling an expensive "natural shield" that actually puts the child at immediate physical risk is unethical. There is no science here—only profit. An ethical doctor prescribes evidence-based treatments, not potentially dangerous, unproven sprays that potentially risk closing off a toddler's airway. ───#primerry #coughcold #immunoboosters #probiotics #herbalspray
2
12
21
686
Debjit B. retweeted
Founders getting ready to do 30 minutes of marketing after 1 year of building features:
121
269
61
2,597
115,766
Same medicine. ₹4 crore a year vs ₹2.5 lakh a year. 💔 A helpless mom came to me with her child, who has Allan-Herndon-Dudley syndrome (MCT8 deficiency). It was very painful to see them so helpless. I tried to find a solution as if it were my own child. I couldn't rest otherwise. Tiratricol wasn't available in India. Dr Santosh and I wrote to doctors. Dr Praveen and others directed us to the Erasmus team, who did the trials for Tiratricol. After many emails, they suggested Egetis Therapeutics might be able to support the child. I teared up. But our happiness was short-lived: Egetis said it could not give us the medicine. I wrote to many journalists. I posted on Instagram and tagged the PM. Someone told me Tiratricol was available in Germany. The mother bought it, paying ₹12 lakh for just one month. That's impossible to keep up for life, for this child or for the five other children with the same condition under Dr Praveen's care at CMC Vellore. Finally, Toufiq, a senior journalist, wrote about it, and it reached the Pharmaceutical Accountability Foundation in the Netherlands. They said Tiratricol can be made available at ₹2.5 lakh per child per year, against ₹4 crore per child per year by Egetis! I raised funds, and the Pharmaceutical Accountability Foundation raised funds. Now 7 children are getting the medicine free of cost for one whole year. (The 7th child was added after my Instagram post.) 🙏 Thank you to Access to Medicines India and Organization for Rare Diseases India for their important role, and to geneticist Dr Kishore, who helped us reach the diagnosis. India can compound the Active Pharmaceutical Ingredient here and make it available to every child with this condition at a reasonable cost. 👉 Please tag @narendramodi @JPNadda @MoHFW_INDIA People of India, you can make it happen! Please share. 🙏 All the paediatricians, endocrinologists, paediatric endocrinologists, and paediatric neurologists, if you are currently caring for a child with confirmed or suspected MCT8 deficiency, please share the details below. Our team may contact you for further clinical information and coordination regarding diagnosis, treatment and access to tiratricol. Please do not upload or share unnecessary patient-identifying information at this stage. docs.google.com/forms/d/e/1F… #AllanHerndonDudley #MCT8Deficiency #Tiratricol #Triac #Emcitate #RareDiseases #RareDiseaseIndia #AccessToMedicines #PharmaHijack #ChildHealth
39
360
37
912
23,059
So we just needed a polished UI and OSS ASR model? Let's create that wrapper. :)
25
With @NVIDIAAI and the @devpost hackathon, I made a meetings feature for my lipi. devpost.com/software/lipi-vo… Please share your feedback. Thanks.
1
38
How???
My first 26 days on 𝕏 wrapped: • 2.9M views across 97 posts • +6400 followers • ~30k median views per post • Most-viewed post: 326k from Distribution 101 article • 26-day posting streak • Topics: Head of Content, content strategy, X growth, creator monetization & related topics • 86/98 posts were value pieces Cool tool @miracleinvxker
1
21
Debjit B. retweeted
On a call, you either type notes and half-listen or listen properly and forget details by tomorrow. I built Lipi Meetings (beta) so you don't have to pick. Local transcript, summary, decisions, and action items. Ask first. ★ github.com/debjit/lipi lipi.debjit.in #buildinpublic
1
15
Disgusted by the profit merging. If you do not have money, you will die a painful death. 73->696
The National Pharmaceutical Pricing Authority has released a list of brands of anti-cancer medicines with their revised Maximum Retail Price (MRP). The revised price list reflects the information submitted by companies up to October 27, 2026.
31
Debjit B. retweeted
This is what PHP2SHELL looks like 👀 Security issues comes everywhere, it only affects you if you do not update on time!
2
2
15
770
On a call, you either type notes and half-listen or listen properly and forget details by tomorrow. I built Lipi Meetings (beta) so you don't have to pick. Local transcript, summary, decisions, and action items. Ask first. ★ github.com/debjit/lipi lipi.debjit.in #buildinpublic
1
15
Lipi 0.1.8 is live with Meetings (beta). Record your mic + system audio on any call. When you stop, you get a transcript, summary, decisions, and action items saved locally on your PC. Tell everyone before you record. ★ github.com/debjit/lipi #buildinpublic
1
1
1
49
Debjit B. retweeted
Filters are dead. 👀 Okay, maybe not. 😂 Why are we making users translate what they want into our filter UI? Just describe what you’re looking for. Then use filters when you want more control. 🎛️ I still want filters. I just don’t think they should be the first thing users see. Sponsored by @typesense. fandf.co/4Ahm926
5
15
3
26
2,818
Debjit B. retweeted
We are launching a PHP Capture the Flag challenge integrated with our Bug Bounty program to test our WAF protections. Security researchers are invited to bypass our defenses and capture the flag to help make the Internet safer. cfl.re/4iqHaQ1
32
124
37
886
257,588
Following our initiative of defending with every possible model, we’re launching OrcaCyber GLM-5.3 Flash Abliterated, one of the most powerful OSS models, for all Orca Cyber Tier users. FREE for 48 hours. Limited compute, first come, first served. orcarouter.ai/models/orca/or…
13
5
7
188
17,091