Transcribe

Our model achieves 98%+ accuracy on clear UGC-style speech, which is consistently higher than general-purpose transcription services on creator content specifically. We fine-tune on creator speech patterns, informal grammar, and the specific vocabulary common in DTC and e-commerce ad scripts — fast delivery, energetic phrasing, colloquial language. Accuracy is lower for very soft speech, overlapping audio, or content with loud background music. We recommend generating audio without heavy background music if you plan to transcribe, as clean speech produces significantly better output than speech mixed with music.

Yes. Our transcription model supports all 74 languages available on the BasedUGC platform. For multilingual content, select the source language before transcription for optimal accuracy, or use auto-detect mode for mixed-language content. Transcription accuracy varies by language — it's highest for English, Spanish, French, German, and Japanese (the most represented in our training data) and slightly lower for lower-resource languages. We continuously add training data for all supported languages. For translated videos generated using our Translate Video tool, transcription automatically runs in the target language without manual configuration.

SRT (SubRip Text — the standard subtitle format compatible with YouTube, Vimeo, Meta, and all major video platforms), VTT (WebVTT — for web-based players and HTML5 video), TXT (plain text, with optional timestamps), and JSON (structured data with word-level start and end times, useful for developers integrating with their own tools or ad platforms). All formats preserve the word-level timing data from the transcription model. For simple caption generation within BasedUGC, you don't need to export at all — the transcript feeds directly into the Captions tool with a single click.

Yes. The transcript opens in a fully editable interface where you can click any word and modify it, delete filler words, correct rare errors, or rewrite segments for cleaner caption output. Changes update the timing data automatically — if you delete a word, the surrounding timing adjusts so there's no gap in the caption output. The edited transcript can be used as the source for Captions, exported in any supported format, or saved to your content library. For SRT export specifically, you can also adjust subtitle grouping — controlling how many words appear in each caption block for optimal readability across different screen sizes.