SKAVIO / JOURNAL
PRODUCT NOTE
Skavio Transcribe turns long audio into usable files inside Telegram
Long recordings do not need to be split into short clips first. Skavio Transcribe accepts audio, video and document uploads in Telegram, processes them through a persistent queue and returns transcription, translation or subtitles in practical file formats.
From recording to document
Send a supported file to Skavio Transcribe + Translate and choose the result you need. A recording can become plain text for quick reading, a DOCX or PDF for sharing, or subtitle files for video work. Translation can be added to the transcription or subtitle workflow.
- TXT, DOCX and PDF transcription outputs
- SRT and VTT subtitle outputs
- Translated transcription and translated subtitles
- Telegram voice, audio, video and document uploads are supported
Long jobs remain visible
Long audio is handled as a queued job rather than an opaque upload-and-wait process. The service keeps extra jobs in a persistent FIFO queue. In the Mini App, queued customers can see their queue position and an approximate estimated-start range, while processing, finalizing and completed states remain visible.
A measured two-hour test
In a measured readiness test, a real two-hour source completed end to end in about 4 minutes 54 seconds. OpenAI transcription itself took about 2 minutes 14 seconds in that run. These are recorded test results, not a promised processing time: actual performance depends on the job and audio quality.
Pricing follows the job
Transcription costs 2 Telegram Stars per started minute. Transcription with translation costs 3 Stars per started minute. Subtitle generation costs 2 Stars per started minute, while translated subtitles cost 3 Stars per started minute. The bot shows the exact Stars price before payment.
- Direct public URLs are accepted up to 500 MiB
- YouTube is currently not supported from the production server because of upstream bot checks
- Third-party websites may block automated access
- Accuracy depends on the quality of the source audio
Recovery is part of the workflow
The queue and recovery system were tested with two real two-hour jobs in an isolated test database using production queue and recovery modules. Both jobs completed, and no duplicate delivery was observed. Supported paid processing failures retain the automatic Telegram Stars refund path.
Long recordings are treated as durable queued jobs, with the work state and approximate start range visible instead of hidden.