Cleanvoice AI
Tool · this deep dive is built from the official pages of the platform itself and of the tools that support it.
Overview
Cleanvoice AI cleans speech recordings and video podcasts automatically: it removes filler words, stutters, background noise, silence and dead air, mouth sounds and breathing, echo and reverberation. Alongside these are declared podcast mixing, the Studio Sound audio enhancer, video podcast editing, transcription with subtitles, summaries and show notes, export of markers and of the timeline into editing programmes, an API and a set of free auxiliary tools. The composition is taken from the pricing page.
About the languages a number is known, but not a list. The transcript page says: "We can transcribe your podcast from over 60 languages" — and the list of those sixty languages is not on it. Groups are named: "English, Latin languages (Italian, French, Spanish etc), Slavic Languages (Russian, Polish etc) and European Languages in general work well". Russian is named directly. Arabic, Hindi and Chinese are not in the list — but that is not a prohibition: the page does not write about them. By plan the languages are not divided: "All plans include all features".
Which platforms it works with
Social networks are not on this list — editing programmes are. The integrations page: "We currently support Adobe Audition, Adobe Premiere, Davinci Resolve, Reaper and Audacity. And also anything which supports.EDL (Edit Decision Lists)". What is exported are markers of the edits made and a timeline of the edits proposed.
The transcript is exported in.SRT and, according to the transcript page, is suitable for YouTube subtitles. There is an API, an API sandbox and an integration with Make.
Direct publishing to social networks and to podcast hosts is not in what was read.
Pricing
Two schemes. The one-off one, Pay as You Go, with credits valid for two years: 5 hours — 11 USD (at 2.20 an hour), 10 hours — 20 USD (at 2.00), 30 hours — 45 USD (at 1.50). The subscription: 10 hours a month — 11 USD (at 1.10 an hour), 30 hours — 30 USD (at 1.00), 100 hours — 90 USD (at 0.90); there is a switch to annual payment.
The currency switches: the tool shows the sums in dollars or in euros, and the title of the pricing page declares the minimum in euros — "Pricing - Start Free, Plans from €10/month". In dollars the minimum subscription is 11 USD a month for 10 hours; the cheapest one-off pack of 5 hours costs the same, but the credits in it live longer.
The Custom Plan — from 200 hours, with separate API endpoints and priority support — is by arrangement. Prices exclude VAT, payment goes through Paddle.
What is free
Thirty minutes of audio and video: "Clean 30 minutes of audio and video for free", no card needed. On the pricing page the tier is called Free Trial — "Try our service for free. No credit card required".
Separately there is the Cleanvoice Startup Program: eligible startups receive 100 hours of credits for the advanced features.
Restrictions
What is billed is the duration of the file, with a one-minute minimum and rounding up: "If you process a 10min 20sec audio file, we bill 11 minutes". Multitrack files are not billed additionally as yet.
Unspent subscription credits carry over, but with a ceiling: "You can roll over up to 3 x your monthly credit budget". Pay as You Go credits are valid for two years. Subscription credits are drawn down first, then Pay as You Go. On cancellation of a subscription its balance is available only until the end of the paid period. On a failed upload nothing is charged.
Storage is short: the source and the processed file are deleted after seven days, and uploads made without an account within four hours. The same terms are repeated in the terms and conditions, clauses 11.2 and 11.11.
Rights to the output
The rights stay with the user. The terms and conditions, clause 7.3: "you also retain all right, title and interest in and to the Results generated through the Services". The licence you grant the tool (clause 7.4) is given exclusively for the provision of the service ordered.
Restrictions on commercial use of the output are not in what was read. Differences by plan are absent too — neither in the terms nor in the price table: the pricing page says directly "All plans include all features".
Do they train on your data
No, and this is written down as a direct prohibition on itself. The terms and conditions, clause 11.7: stored content is not used "for training artificial intelligence models or for any purpose unrelated to the provision of the Services".
A mechanism for opting out is not in the text, and this is not a gap but a consequence: since training is forbidden by the terms, there is nothing to opt out of.
How you earn with it
On commercial use the answer is definite: the rights to the output the terms leave with the user, restrictions they do not impose, and by plan they divide nothing.
What it will not do
It will not publish a clip or an episode for you: direct publishing to social networks and to podcast hosts is not in what was read.
It will not clone someone else's voice: no such function is described on the pages read, and there are no rules about cloning in the terms.
It will not check the consent of the people recorded for you, but it will demand it. Clause 7.9.1: the user has "all rights, licenses, permissions and legal bases necessary to upload, process and use the Content". Clause 7.9.4: if the recording contains personal data, the user has "all necessary legal grounds, notices, authorizations and, where required, consents for such processing". A description of any technical verification of consent is absent; a direct ban on processing the voice of a public figure is absent too.
It will not take what is forbidden. Clause 9.3 names unlawful, defamatory, abusive, threatening, discriminatory and sexually explicit content, the distribution of malware, circumvention of technical restrictions and security measures, reverse engineering of the platform, competitive benchmarking and model extraction.
It will not keep files for long: seven days with an account, four hours without one.
It will not name all sixty languages: the number is declared, the list is not published.
It will not mark the processed file as having passed through AI: neither a watermark, nor an audio marker, nor metadata, nor C2PA is mentioned in the terms or on the pricing page.
Verified data
The checked data this deep dive rests on.
About the platform
Automatic cleanup of speech recordings and video podcasts: removing filler words, stutters, background noise, silence and dead air, mouth sounds and breaths, echo and reverb; podcast mixing, Studio Sound (audio enhancer), video podcast editing, transcription with subtitles, summaries and shownotes, export of markers and a timeline to editing software, an API and a set of free helper tools
The feature set was checked against the Access to all features block on the pricing page and against the cleanvoice.ai footer
source, checked 2026-08-07
Transcription: «We can transcribe your podcast from over 60 languages» (verbatim); a full list of the 60+ languages is not on the page. Quality is described by groups: «English, Latin languages (Italian, French, Spanish etc), Slavic Languages (Russian, Polish etc) and European Languages in general work well» (verbatim) — Russian is named directly; Arabic, Hindi and Chinese are NOT mentioned in what was read. It does not differ by plan: «All plans include all features»
Marked partial: the number of languages is claimed but the list itself is not published; the absence of Arabic, Hindi and Chinese in the text is silence of the source, not a prohibition
source, checked 2026-08-07
Platforms
Export of edits to editing software: «We currently support Adobe Audition, Adobe Premiere, Davinci Resolve, Reaper and Audacity. And also anything which supports .EDL (Edit Decision Lists)» (verbatim); what is exported is markers of the edits made and a timeline of the proposed edits. The transcript is exported as .SRT and suits YouTube subtitles. There is an API, an API sandbox and a Make integration. Direct publishing to social networks and podcast hosts is not in what was read
The option of using the transcript as YouTube subtitles comes from https://cleanvoice.ai/podcast-transcript/
source, checked 2026-08-07
Pricing
Free: 30 minutes of audio and video — «Clean 30 minutes of audio and video for free» (verbatim), no card needed; on the pricing page the tier is called Free Trial — «Try our service for free. No credit card required». Separately there is the Cleanvoice Startup Program: eligible startups receive 100 hours of credits for the advanced features
The 30-minute allowance is repeated on several of the service's product pages; the conditions of the startup programme are in the FAQ block of the pricing page
source, checked 2026-08-07
Pay as You Go (credits valid for 2 years): 5 hours — $11 ($2.20 an hour), 10 hours — $20 ($2.00), 30 hours — $45 ($1.50). Subscription: 10 hours a month — $11 ($1.10 an hour), 30 hours — $30 ($1.00), 100 hours a month — $90 ($0.90); yearly billing can be switched on. The heading of the pricing page states the minimum in euros: «Pricing - Start Free, Plans from €10/month». A Custom Plan (200+ hours, dedicated API endpoints, priority support) is by arrangement. Prices exclude VAT, payment through Paddle
The amounts are taken from the pricing tables; the service shows them in USD or EUR depending on the switch, and recorded here are the USD values plus the EUR minimum from the page heading
source, checked 2026-08-07
11 USD a month — a subscription for 10 hours of processing a month. In euros the minimum tier is stated in the heading of the pricing page as 10 EUR a month. The cheapest one-time purchase is a 5-hour pack for 11 USD
The monthly subscription for 10 hours and the one-off 5-hour pack cost the same but differ in how long the credits last
source, checked 2026-08-07
Limits and restrictions
The duration of the file is charged, with a 1-minute minimum, rounded up: «If you process a 10min 20sec audio file, we bill 11 minutes» (verbatim). Multitrack files are not charged extra for now. Unspent subscription credits carry over, but with a ceiling: «You can roll over up to 3 x your monthly credit budget» (verbatim). Pay as You Go credits are valid for 2 years. Subscription credits are spent first, then Pay as You Go. If the subscription is cancelled, its unspent credits are available only until the end of the paid period. A failed upload is not charged. Storage: the source and processed files are deleted after 7 days, uploads without an account within 4 hours
The retention periods are repeated in the Terms & Conditions at https://cleanvoice.ai/toc/ (clauses 11.2 and 11.11)
source, checked 2026-08-07
Restrictions
WHOSE VOICE: THERE ARE NO rules about CLONING someone else's voice in what was read — no such feature of the service is described on the pages read. Requirements for UPLOADED recordings, on the other hand, are stated directly: clause 7.9.1 — the user holds «all rights, licenses, permissions and legal bases necessary to upload, process and use the Content» (verbatim); clause 7.9.4 — if the recording contains personal data, the user has «all necessary legal grounds, notices, authorizations and, where required, consents for such processing» (verbatim), that is, consent from the people recorded is required where the law requires it. There is no outright ban on processing the voice of a public figure; nor is any technical check of consent described. FORBIDDEN TOPICS (clause 9.3): unlawful, defamatory, abusive, threatening, discriminatory and sexually explicit content, distributing malware, circumventing technical restrictions and security measures, reverse engineering the platform, competitive benchmarking and model extraction
The consent requirement comes through the personal data regime rather than through a separate rule about voice
source, checked 2026-08-07
Legal
Rights to the output stay with the user. Terms & Conditions clause 7.3 verbatim: «you also retain all right, title and interest in and to the Results generated through the Services». The licence to the service (clause 7.4) is granted solely to deliver the service ordered. There are no restrictions on commercial use of the output in what was read. THERE IS NO DIFFERENCE BY PLAN either in the terms or in the pricing: the pricing page states outright «All plans include all features»
Checked both in the terms and in the pricing table: no functional or legal division between tiers was found
source, checked 2026-08-07
NO, training is expressly forbidden by the service itself. Terms & Conditions clause 11.7: stored content is not used «for training artificial intelligence models or for any purpose unrelated to the provision of the Services» (verbatim)
One of the few sets of terms in this selection where training on user audio is forbidden in explicit wording
source, checked 2026-08-07
Catalogue section: all similar See also: catalogue index · scheduler comparison · find by situation · platform restrictions