UniScribe

Free includes 120 minutes of transcription a month, but each file can be at most 30 minutes long, only one file uploads at a time and no more than 3 files can be transcribed per day. Free uses the basic transcription model, which the pricing page labels standard accuracy, and uploaded media is kept for 30 days.

The free plan card lists translation, all six export formats and limited AI insights. YouTube transcription, speaker identification, custom vocabulary and the API appear only on paid plans. A 90-minute lecture or a two-speaker interview that needs names attached to each line is therefore a paid-plan job.

Subscriptions reset their minutes each billing cycle, while one-time packages are a single payment for a fixed number of minutes: Lite and Plus last 90 days and Max lasts 180 days. Packages include the premium model, speaker identification and exports, but the package cards do not list API access or the Claude and ChatGPT integration.

If you hold several balances, minutes are deducted from whichever expires first, and an exhausted package falls back to any other plan or to Free. Editorial assessment: a package suits one batch of recordings within 90 days, while Basic is cheaper per minute for anyone who transcribes every month.

Subscribers can buy add-on minutes from Settings, then Usage. The packs are fixed by plan: $10 for 500 minutes on Basic, $15 for 1,000 on Standard and $20 for 3,000 on Pro. Add-ons last 90 days, can be bought several times and are limited to subscription and lifetime-deal accounts.

Plan minutes reset at the start of each cycle and do not roll over. Upgrading from Free starts the paid billing cycle on the day of the upgrade and charges the card immediately, according to the terms. Choose the tier from your typical month, and use add-ons only for occasional peaks.

The pricing FAQ says a monthly subscription can be refunded for the current period if no transcription minutes were used, and an annual plan can be refunded pro rata for the remaining time, by emailing hi@uniscribe.co. The terms, last updated on October 9, 2024, state instead that all purchases are non-refundable.

Because the two pages disagree, keep the confirmation email and ask support before relying on a refund. Cancelling keeps the subscription active until the end of the billing period, after which the account moves to Free. Deleting the account ends access to all content immediately.

The privacy policy lists Microsoft Azure in the United States and Cloudflare for hosting, and OpenAI and Replicate among the AI providers behind certain features. Staff review a specific transcript only with explicit permission, for example when you rate its quality. The security page says files are encrypted at rest with AES-256 and are not used for AI training without consent.

Recordings often contain other people, and the policy asks you to obtain their permission before uploading. Deletion wording varies by page: within 24 hours in the privacy policy and within 30 days on the security page. Export what you need, then delete the transcript and the source file.

The REST API, still labelled beta, uploads files, starts transcriptions, returns transcripts and sends webhooks. It requires an active subscription or a lifetime deal; Free users and one-time package buyers cannot create API keys. The documentation allows files up to 5 GB and 10 hours, while plan cards state 5 hours.

A separate read-only MCP server lets Claude, ChatGPT or other MCP clients search and cite transcripts. Under the privacy policy, a connected app can read transcripts and create new transcriptions within the permissions you grant, but it cannot edit or delete content. Connections can be revoked in Settings, under Connected Apps.

1. Note the remaining monthly minutes before uploading.
2. Transcribe the file on the basic model in the recording language.
3. Export SRT and TXT and open both outside UniScribe.
4. Compare the five technical terms and the noisy passage with the recording.
5. Record the minutes deducted and the time from upload to finished transcript.

Accept an SRT whose cues stay in sync across the whole file and a TXT with the five terms correct or easy to fix. Reject missing sections, subtitles that drift by more than a second or a deduction larger than the file length. Leave the observed result blank until the exercise is run.

To edit a recording by its transcript, compare Descript. For clean audio before transcription, see Adobe Podcast; for live meeting notes rather than uploaded files, see Krisp.