Portuguese Transcription: Video & Audio to Text

How Portuguese speech transcribes

Brazilian and European Portuguese both transcribe well. European Portuguese, with its reduced vowels, is slightly harder for speech recognition when spoken fast.

What a Portuguese transcript looks like

Portuguese transcripts include accents and the cedilla (ç). Spelling follows the variety spoken, so Brazilian and European pronunciation can lead to slightly different spellings of the same word — useful to know when you search a transcript. Colloquial forms common in Brazilian speech, such as 'tá' for 'está' or 'a gente' for 'we', are often written as spoken.

Examples

Costs: 58 credits (5.8¢) and 19 credits (1.9¢), including a translated transcript if you select another language. Audio files go in the audio to text workspace.

Which spelling norm will you see?

The 1990 Orthographic Agreement brought Brazilian and European spelling closer, but differences remain: facto and fato, receção and recepção, António and Antônio, económico and econômico. The transcript's spelling may not match your audience, and may not even match the speaker's country. If you publish for Portugal, search for the common Brazilian forms, such as ô accents and the -pç- and -ct- clusters Portugal dropped, and do the reverse for Brazil. A short list of the words that recur in your subject covers most cases. Month names and days of the week are another quick check: Portugal writes them in lower case (janeiro, segunda-feira), as Brazil does.

Clitic pronouns and hyphens

European Portuguese places object pronouns after the verb with a hyphen (disse-me, vou-te dizer, dá-lho), while Brazilian speech usually puts them before it (me disse, vou te dizer). Speech carries no hyphens, so check that a Lisbon speaker's enclitic pronouns are hyphenated in the transcript, and that formal future forms with the pronoun in the middle (dir-lhe-ei, far-se-á) are written in full. These are quick to find with search once you know to look for them.

Verb forms give the variety away too. Brazilians say estou fazendo where speakers in Portugal say estou a fazer. Both are correct, so don't standardise them in quotes; they are also a fast way to confirm which variety a speaker uses before you choose a spelling norm.

Meia, mil and decimal commas

Brazilians reading out digits often say meia (short for meia dúzia, half a dozen) for six, so a phone number or order code read as "três meia nove" means 369. Check that the transcript wrote the digit and didn't leave meia as a word or turn it into "half". Portuguese uses a decimal comma and a full stop or space for thousands (1.500,75 or 1 500,75), currency is written R$ in Brazil and with € after the amount in Portugal, and large numbers are spoken as mil and milhão. Search for figures in budgets and price lists and listen to each one.

Recordings from Angola, Mozambique, Cape Verde and Timor-Leste

African and Asian varieties of Portuguese follow European spelling, but vocabulary, place names and personal names often come from Kimbundu, Umbundu, Emakhuwa, Changana, Tetum and other local languages. The transcript will spell these by sound using Portuguese conventions, so give whoever reviews it a list of names and terms. In Cape Verde and Guinea-Bissau, conversation may switch into Kriolu, which isn't one of the 21 languages; expect those passages to come out as approximate Portuguese, and mark them for a fluent speaker. The article on speech recognition and accents explains why less common varieties tend to produce more errors.

When a short clip might be read as Spanish

Language detection is automatic. Portuguese and Spanish share a lot of vocabulary, so on a very short clip, or one that opens with a name, a number or music, confirm that the transcript is actually Portuguese. Spanish spellings such as ñ or -ción where the speaker said -ção mean the wrong language was detected; trimming the intro so the file opens with clear speech is worth trying before you rerun it. Spoken language identification explains how detection decides.

Example: a Lisbon interview podcast

Transcripts don't label speakers, so with two hosts the producer adds names at each change of voice. The interview transcription page covers that editing pass, and M4A to text covers the file type.

Using the tool on this page

• The spoken language is detected automatically; you select the output language — the spoken language itself for an untranslated transcript, or another language for a translation. Files can be up to 2 hours long.
• Only the audio is sent for processing — the video stays on your device — and audio and text are deleted within 30 minutes of delivery (how files are protected).
• You can download a timestamped transcript, subtitles (SRT and VTT) and plain text. Every output, step and limit is explained on the video to text page.