DeVoice
DeVoice is a browser-based AI transcription platform converting audio, video or pasted YouTube links into editable text with speaker labels, timestamps, AI summaries and TXT, SRT, VTT, DOCX or CSV exports. A permanent free plan is included.
What is DeVoice?
DeVoice is a web-based AI transcription platform published by Luminous Swallow Technology Co., Limited. Its own definition is direct: an AI-powered transcription platform that converts audio and video into accurate text. Everything happens in the browser, with nothing to install, and the site says it runs on Windows, macOS, iOS and Android alike.
The product covers two families of functions. The first is transcription: audio files, video files, podcasts, YouTube links and meeting recordings become editable text with automatic speaker labels, per-line timestamps, filler-word removal and an optional AI summary. The second is audio processing: one-click background-noise removal and voice isolation or enhancement. Accepted inputs are mp3, wav, m4a, mp4, mov, webm, mkv, flac, aac, ogg, aiff and avi, up to 500 MB on the free plan and 5 GB on Basic and Pro. Outputs come as TXT, SRT, VTT, DOCX or CSV.
DeVoice advertises 95.2% accuracy on clear speech, and says it holds up across accents, fast delivery and light background noise. Language detection is automatic. Language coverage, however, is stated inconsistently: 100+ languages on the home page, the podcast page and the YouTube page, 50+ on the audio-to-text and video-to-text pages, and over 98 languages elsewhere. The interface itself is localised into eleven languages, though the legal pages are never translated.
Three use cases are pushed on the home page: meeting notes, content repurposing and accessibility. Meeting support works from the audio only, so Zoom, Google Meet, Teams and Webex recordings all produce the same notes, which makes them recording sources rather than integrations. API access is sold inside the Pro plan, yet no API documentation page exists anywhere on the site.
Two caveats belong in any honest description. The terms of use and the privacy policy, both effective 10 April 2025, still describe a vocal-remover and audio-separation product rather than transcription: they were never updated after the repositioning. And while the blog runs from May 2025 through July 2026 and shows a live site, its single comparison article recommends a competitor and never mentions DeVoice at all.
What it does
- Transcribe audio and video files into editable, timestamped text
- Transcribe a YouTube video, Shorts included, by pasting its URL, with no download step
- Identify and label each speaker automatically
- Generate an AI summary listing decisions, action items and owners
- Produce SRT or VTT subtitle files for a finished video
- Export the result as TXT, SRT, VTT, DOCX or CSV
- Remove background noise, then isolate and enhance a voice
When to use DeVoice / When not to
A quick filter to help you decide if DeVoice is the right fit.
When to use DeVoice
- Podcasters and content creators turning a finished episode into show notes, blog drafts and social posts
- Remote and hybrid teams that record their meetings and want a written record with speaker labels and clear who-said-what attribution
- Video makers and social editors who need SRT or VTT subtitle files ready to drop onto a finished video
- Journalists, researchers, lecturers and students transcribing interviews, courses and webinars from an existing recording
- Solo users and small teams on a tight budget: a genuinely usable free plan, then USD 4 to 5 per month, and up to 3,000 upload minutes a month on Pro
When not to use DeVoice
- Anyone who needs live capture: DeVoice always starts from a finished recording, never from a stream in progress. The site's own blog points readers to Otter.ai for real-time note-taking
- Work where an error is expensive, such as legal, medical or press contexts: the site's own benchmark credits human transcription with 2.9% WER against 4.8% for the best AI tool it tested, and sends that reader to Rev.com
- Organisations with a compliance checklist: no data processing agreement, no subprocessor list and no SOC 2 or ISO 27001 certification is published
- Buyers bound by data-residency rules: no hosting country or region is ever named
- Editors expecting a full audio or video suite: apart from noise removal and voice isolation there is no editing, and there is no native mobile app
How to use DeVoice
A typical end-to-end flow, from setup to results.
- Open the tool in any browser: there is nothing to install and no technical setting to configure
- Drag and drop your audio or video file, or use the upload button, or paste a YouTube URL instead, Shorts included
- Sign in with Google if needed: the product pages allow a first transcript without an account, but quotas and credits require one
- Let the AI detect the speech, transcribe it, separate the speakers, place the timestamps and strip the filler words
- Read the result back and edit the transcript inline where the machine got it wrong
- Generate an AI summary from the transcript if you need decisions and action items rather than the full text
- Copy the text or export it as TXT, SRT, VTT, DOCX or CSV
- For noise removal, follow the dedicated four-step flow: add the file, upload it, separate the audio, then compare before and after before downloading
- Feed the tool the cleanest audio you have, since the site notes that source quality drives accuracy
Pros & Cons
Pros
- A free plan that is genuinely usable rather than a demo: full transcription, speaker labels, AI summary and every export format, with no credit card
- A very low paid entry point at USD 4 per month on annual billing, invoiced USD 48 a year, or USD 5 per month without commitment
- Readable, public pricing expressed in minutes rather than in opaque credits
- Five export formats including SRT and VTT, where several competitors stop at TXT and DOCX according to the comparison table published on the site's own blog
- No training on customer content, written in plain terms in the privacy policy, with automatic file deletion after processing
- Native YouTube input: paste a link and skip the downloader step entirely
- Nothing to install, works on any device and any operating system, with the interface localised into eleven languages
Cons
- The site contradicts itself on figures. Language coverage is announced as 100+, 50+ and 98 depending on the page, and the free tier as 3 transcriptions a day on the product pages, 30 transcription minutes a month on the pricing page, or 1 audio processing a day in the terms of use
- The terms of use and privacy policy, both effective 10 April 2025, still describe a vocal-separation product rather than transcription: the legal documents were never updated after the repositioning
- An internal upgrade modal still advertises a price starting at 9.99 USD per month, an amount that matches none of the four published tiers of 4, 5, 19 and 29 USD
- Nothing is disclosed on the company side: no postal address in any of the usual places, no named hosting country, no data processing agreement, no subprocessor list, no SOC 2 or ISO 27001 certification, no Article 27 EU representative, no DPO, and GDPR compliance is never claimed
- Governing law is left unnamed, described only as the laws of the jurisdiction in which DeVoice operates, and a single Cloudflare-obfuscated email address handles support, legal, GDPR, DMCA and refunds alike
- Credits are non-refundable and purchases are final by principle, while API access is sold on the Pro plan with no public documentation to integrate against
- The pricing page is linked from no menu and no footer, reachable only through the sitemap or a direct address; there is no live capture; and the site's only comparison article recommends AudioTranscription.io without ever mentioning DeVoice
Pricing & Plans
DeVoice operates on a freemium model. A free plan is available permanently and without a credit card: it is a standing tier, not a time-limited trial. The lowest paid entry point is the Basic plan at USD 4.00 per month on annual billing, invoiced USD 48.00 per year; the same plan costs USD 5.00 per month without commitment, the annual view being advertised at 20% off and displayed by default on the site. The Pro plan is USD 19.00 per month on annual billing, invoiced USD 228.00 per year, or USD 29.00 per month without commitment, advertised at 34% off. One-time credit packs may be purchased in addition, including by existing subscribers, but no pack price is published on the site. Credit consumption is metered at roughly 2 credits per minute for an uploaded transcription, 1 credit per minute for a captioned YouTube video and 2 for an uncaptioned one, 4 credits per minute for Basic denoising and 20 for Pro denoising. Subscriptions renew automatically, monthly or annually, and may be cancelled at any time, with effect at the end of the current cycle and no pro-rata refund. All purchases are final and non-refundable, with two exceptions to be claimed within seven days: a duplicate charge and a total service failure. A chargeback may lead to account suspension and the loss of remaining credits. No enterprise plan and no quote-based pricing is published.
- full transcription
- standard queue
- speaker labels
- AI summary and all five export formats
- 500 MB maximum per file
- no credit card required. Comparison-table quotas are 30 minutes of captioned YouTube
- 15 minutes of uncaptioned YouTube
- 15 minutes of upload and 5 minutes of denoising
- 600 credits a month
- 600 minutes of captioned YouTube
- 300 minutes of uncaptioned YouTube and 300 minutes of upload per month
- 150 minutes of Basic denoising and 30 minutes of Pro denoising
- priority queue
- 5 GB maximum per file
- batch processing and email support
- 6
- 000 credits a month
- 6
- 000 minutes of captioned YouTube
- 3
- 000 minutes of uncaptioned YouTube and 3
- 000 minutes of upload per month
- 1
- a dedicated tab exists on the pricing page
- but no pack price is rendered in the page's static HTML
Data, GDPR & hosting
A consolidated view of how DeVoice handles your data.
GDPR overview
DeVoice never claims GDPR compliance anywhere on the site, and never denies it either. Section 6 of the privacy policy acknowledges rights depending on your location, naming GDPR and CCPA: access, deletion, rectification, objection and withdrawal of consent, exercised by writing to service@devoice.io, with a stated 30-day response. Beyond that the file is thin. No legal basis is named, no international transfer mechanism or standard contractual clauses, no data processing agreement offered or even mentioned, no subprocessor list, no Article 27 EU representative and no DPO. Cookies are described as essential plus analytics, with Google Analytics named and browser settings as the only opt-out, and no consent banner is documented. The minimum age is 13. The publisher is a non-EU company that publishes no postal address and names no hosting country. European users are told they have rights, but nothing contractual stands behind them.
Who owns the data?
Ownership stays with you. The privacy policy, effective 10 April 2025, states that your content remains your property and that DeVoice claims no ownership of any audio or media you upload. Section 5 of the terms of use repeats it twice over: you retain all rights to the content you upload, and DeVoice does not own your media. The publisher processes files only to deliver the service, and the transcripts, subtitles, summaries and cleaned audio you get back are yours. The counterpart is that sections 4 and 5 of the terms place responsibility for copyright entirely on you: you must hold the rights to every file you submit.
Reuse rights
You can reuse everything DeVoice gives back without asking permission. Transcripts, subtitle files, summaries and cleaned audio are yours to edit, publish, monetise or repurpose, provided you hold the rights to the source recording. On the publisher's side, files are processed temporarily on its servers and then deleted, and the privacy policy is unusually explicit: content is not used for training, analysis or any secondary purpose, and data is not used for targeted advertising, profiling or resale to third parties. The product pages repeat the same commitment. No account is required for the basic service, and personal data is collected only if you write in, in which case it is your email address, name and message content. Technical logs cover IP address, browser, device, operating system, timestamps, pages viewed and referrer, and the policy lists seven purposes for them: operating the service, processing files, troubleshooting, improving the experience, preventing abuse, replying to you and meeting legal obligations. Sharing is limited to three categories of providers, namely cloud storage, analytics and security or DDoS protection. None of them is named, apart from Google Analytics cited as an example.
Data retention & training
Hosting summary
No hosting country and no hosting region is ever named, neither in the privacy policy nor in the terms of use nor anywhere else on the site. The privacy policy says only that files are processed temporarily on secure servers, and refers to unnamed cloud storage services used for that temporary processing. The publisher, Luminous Swallow Technology Co., Limited, publishes no postal address either, so no jurisdiction can be inferred from the company side. Stated security measures are generic: HTTPS, encryption, firewalls and secure servers, with the policy adding that no method is 100% secure and that using DeVoice means accepting that risk. Product pages repeat that files are encrypted in transit and deleted after processing. No SOC 2, ISO 27001 or equivalent certification is published. One technical caveat matters: the domain resolves to a Cloudflare anycast address, which identifies a network edge node and says nothing about where the data actually sits, so it must not be read as a hosting location. In practice, anyone with a data-residency requirement has nothing here to rely on.
Things to keep in mind
Risks and trade-offs to weigh before adopting DeVoice.
- The legal documents do not describe the product you are actually using. The terms of use and privacy policy, effective 10 April 2025, cover a vocal-separation service, not transcription, so the written commitments do not explicitly cover the service you would be paying for
- The free tier's limits contradict each other from page to page, at 3 transcriptions a day, 30 minutes a month or 1 audio processing a day, and the terms tie credits to the browser session until accounts are introduced. Unused credits are non-refundable, so buy only what you will spend
- No postal address, no named hosting country, no data processing agreement, no subprocessor list and no certification. If your recordings carry personal data, interviews, patient files or client information, you are sending them somewhere you cannot locate under terms you cannot audit
- No Article 27 EU representative and no DPO, even though the interface is localised into eight European languages and the site plainly addresses a European audience. Governing law is never named, which makes any dispute hard to place
- A single email address, published only behind Cloudflare obfuscation, serves support, legal, GDPR, DMCA and refunds. There is no second channel if it stops answering, and the pricing page is not linked from any menu
- An internal modal advertises a price that exists in no published tier, so check the amount actually charged at checkout, and remember that a chargeback can suspend the account and forfeit remaining credits
- Treat 95.2% accuracy as a claim, not a guarantee. Machine transcripts read fluently while quietly mangling names, figures and negations, and a generated summary is a convenience, not evidence. Proofread before quoting, publishing or acting on a transcript, and do not let the summary replace having listened to the recording yourself
Setup & Integrations
Technical difficulty
Very low. Nothing to install, no software, no technical settings: the whole tool runs in the browser, and the site states that no software installation or technical setup is required. On the product pages a first transcript can be produced without an account or a card; an account, opened through Google sign-in, becomes necessary for quotas and credits. The only useful skill is supplying decent audio, since source quality drives accuracy. The one exception is the Pro plan's API access: with no public documentation, the integration effort cannot be assessed.
Deployment
Integrations
Supported languages
Behind DeVoice
Social
Resources
All the official URLs gathered for verification and reference.
Alternatives
Tools that compete with or complement DeVoice.
Frequently asked questions
Is DeVoice free?
Which file formats can I upload?
What can I export?
Can I transcribe a YouTube video?
How accurate is the transcription?
How many languages are supported?
Are my files used to train AI models?
How long are my files kept?
Is there a mobile app or an API?
Can I get a refund?
Should you pick DeVoice?
DeVoice is a low-cost, generalist AI transcription platform built around volume and export. Its strongest card is a free plan that is not crippled on features: full transcription, speaker labels, AI summaries and all five export formats, with no credit card. That is rare at this price level, and it sits behind a readable pricing grid expressed in minutes rather than credits, with a paid entry point at USD 4 per month on annual billing and USD 5 per month without commitment.
The reservations are real, and they are the publisher's own doing. The terms of use and the privacy policy, both effective 10 April 2025, still describe a vocal-remover product rather than transcription: the legal commitments lag behind the service they are supposed to cover. Figures contradict each other from page to page on language coverage, on free-tier quotas, even on price, where an internal modal still shows an amount matching no published tier. On the company side there is no postal address, no named hosting country, no data processing agreement and no certification.
The practical verdict follows from that split. For an individual or a small team wanting clean, exportable transcripts cheaply, DeVoice is a sound choice: the privacy policy rules out training on your content, files are deleted after processing, and the outputs stay yours. For a regulated context, anything carrying a data-residency or audit requirement, the paperwork simply is not there. And for note-taking during a meeting in progress, the tool does not apply at all, since it always starts from a finished recording. Read the free plan as what it is, a permanent tier, and test it on your own audio before paying: that is the fastest way to find out whether 95.2% accuracy holds on your material.
- Choosing a selection results in a full page refresh.
- Opens in a new window.