5 AI models
Choose Fast, Accurate, Balanced, Enhanced or Standard. Each shows its price per minute and file limit before you start.
Convert audio and video into accurate text, AI summaries, notes and chat.
Upload one or more files (up to 100 at once), record with your microphone or paste a link.
Select a model for speed or accuracy. Choose the recording language or let the model detect it. You see the price in credits before you start.
Listen and read along, generate AI summaries or meeting notes in one click, ask questions in the AI chat, and download TXT, DOCX, PDF, SRT or VTT.
Convert audio and video into accurate text, summaries, notes and actionable insights.
Choose from five transcription models, chat with your recordings and export in any format. Upload audio (MP3, WAV, M4A, OGG, FLAC and more) or video (MP4, MOV, MKV, AVI, WEBM) up to 3 hours and 1 GB, record your voice or paste a YouTube, Google Drive or direct file link. Everything runs in the browser on a computer or a phone — nothing to install.
One recording with two speakers, exported to four formats: plain text, a document and two subtitle files. Your file gets the same structure.
[00:01] Speaker 1: [00:05] Speaker 2:
Speaker 2 00:05
1 00:00:01,200 --> 00:00:04,500 2 00:00:05,100 --> 00:00:08,300
WEBVTT 00:00:01.200 --> 00:00:04.500 <v Speaker 1> 00:00:05.100 --> 00:00:08.300 <v Speaker 2>
Subtitle lines are 42 characters by default. You can change this before you download.
Go beyond a raw transcript. Generate structured reports in one click or ask questions in the built-in AI chat — every answer is based on your recording, with clickable timestamps.
Get the main topic, key decisions, numbers, dates and responsible people in 3 to 5 sentences.
Structured notes grouped by topic: context, discussion points, arguments, decisions and open questions.
Confirmed tasks with owner, deadline and exact quote [MM:SS], separated from items that still need clarification.
Formal meeting minutes where every statement is backed by a verbatim quote, speaker name and timestamp.
The main ideas and conclusions from a lecture, podcast, interview or webinar, each linked to a timestamp.
Topic-based video chapters starting at 0:00 with clickable timestamps, ready to paste into a YouTube description.
AI chat with your recording: use quick prompts ("What is it about?", "Key takeaways", "Quick summary", "What needs to be done?") or ask your own question in any language. Click any [MM:SS] timestamp in the answer to hear that moment, and download any report as PDF, DOCX or TXT.
Choose Fast, Accurate, Balanced, Enhanced or Standard. Each shows its price per minute and file limit before you start.
Each speaker gets a label. Rename a speaker once and the name changes in the whole transcript. Move a segment to another speaker.
Click a word to play the audio from that point. The current word is highlighted during playback.
With regional variants of English, Spanish, Arabic and other languages. Most models can detect the language automatically.
Record in the browser, pause and resume, then send the finished recording. For voice notes, use voice to text in the browser.
Upload a video file or paste a YouTube or Google Drive link. The audio is extracted automatically.
For files longer than 30 minutes, transcribe only the first minute to check the quality. Then transcribe the whole file.
Not happy with the result? Run it again with another model or language. You see the price before you confirm.
Batch work: up to 100 files per upload and up to 20 files per export — enough for an interview series or a semester of lectures.
Each model has its own strengths and price. Not sure which to pick? Start with Fast, or transcribe the first minute with two models and compare.
| Model | Best for | Accuracy | Speed | Languages | Detects language | File limit | Price |
|---|---|---|---|---|---|---|---|
| FastMicrosoft Fast | The quickest results and the longest language list. Costs twice as much per minute. | 76 | Yes | 3 h / 1 GB | 278 credits/min | ||
| AccurateMAI-Transcribe-2 | High accuracy for interviews, lectures and meetings. Shorter file limit. | 57 | Yes | 2 h / 300 MB | 139 credits/min | ||
| BalancedGrok Voice Transcribe 2.0 | High accuracy at a medium speed. Suits long recordings up to 3 hours. | 25 | Yes | 3 h / 1 GB | 139 credits/min | ||
| EnhancedMicrosoft LLM Speech | Speech recognition with a built-in language model for the most common languages. | — | 24 | Yes | 3 h / 1 GB | 139 credits/min | |
| StandardWhisper (Microsoft Azure) | Works with the full language list. Select the language yourself. Slower than other models. | 76 | Manual | 3 h / 1 GB | 139 credits/min |
Prices are in credits from your SpeechGen balance. The exact amount for your file is shown before you start.
70+ languages · 130+ regional variants · automatic language detection
English, Spanish, Portuguese, French, German, Italian, Japanese, Korean, Hindi, Arabic, Turkish and 60+ more. Pick the recording language or let the model detect it — speech to text in Spanish, German or Arabic works the same way. The list depends on the model: Fast and Standard support the most languages.
Get interview quotes with the speaker and exact time, or generate minutes backed by verbatim quotes.
Turn lectures and seminars into structured study notes, key takeaways and searchable text.
Get show notes, key takeaways, YouTube chapters with timestamps and SRT/VTT subtitles from one upload.
Turn calls and meetings into a brief summary, a table of decisions and action items, or full meeting minutes.
Download SRT or VTT and adjust line length. Working from a video file? Use video transcription.
Start from an accurate source text in the original language.
Transcription is paid with credits from your SpeechGen balance — the same balance you use for text to speech. Most models cost 139 credits per minute of audio; Fast costs 278. The price for your file is shown before you start, and if a transcription fails, the credits are refunded.
Accuracy depends on the audio and the model. Clear speech from one or two speakers gives the best results. Background noise, music, overlapping voices, strong accents and phone-quality audio cause more mistakes. Always check names and numbers before you publish.
It depends on the file length, the model and the current queue. You can close the page: the transcript is saved in your files and appears when it is ready.
Yes. Audio and video files up to 1 GB and 3 hours (the Accurate model: up to 300 MB and 2 hours). Common formats: MP3, WAV, M4A, OGG, FLAC, AAC, MP4, MOV, MKV, AVI, WEBM. You can also paste a YouTube link, a public Google Drive link or a direct link to a file.
More than 70 languages, many with regional variants. Most models detect the language automatically; with the Standard model, you select it yourself. If your recording has a rare language, transcribe the first minute to check the quality.
You pay with credits from your SpeechGen balance, per minute of audio. The price depends on the model and is shown before you start.
You can rename speakers, move a segment to another speaker and delete segments. To correct the words, download DOCX or TXT and edit the file in your text editor.
With an account — 90 days, without an account — 1 day. You can delete a file earlier. Only you can open the result page; to share a transcript, download it and send the file.
No. It works in the browser on a computer or a phone. Upload a file, record your voice or paste a link.
Start with Fast for most recordings. For interviews and lectures where every word matters, try Accurate. If your language is not in the list, use Standard and select the language yourself.
After full transcription, open the AI tab to generate a brief summary, detailed notes, action items, meeting minutes with quotes, key takeaways or YouTube chapters. You can also ask questions in the AI chat — every answer is based on your transcript with clickable [MM:SS] timestamps, and reports can be downloaded as PDF, DOCX or TXT.
Upload a file, record your voice or paste a link. You see the price before you start.