Skip to main content

Multilingual Transcription Prompts

8 copy-ready AI prompts for translating, summarizing, and analyzing non-English and mixed-language transcripts. Works with ChatGPT, Claude, Gemini, and all major AI tools.

BrassTranscripts supports 102 languages — the full FLEURS benchmark language set — so researchers, international teams, and content producers can upload audio in Dutch, Spanish, Mandarin, Arabic, or dozens of other languages and receive a speaker-labeled transcript within minutes. These 8 multilingual prompts turn that raw output into translated documents, executive briefs, code-switching timelines, and localization-ready subtitle files using any AI tool you already have.

102
Languages Supported
8
Multilingual Prompts
100%
AI Tool Compatible
Copy
Ready Format

Download the full prompt library

All 8 multilingual prompts plus 122+ others — available as Markdown and YAML files.

How to use these prompts

1

Get your transcript

Upload your audio to BrassTranscripts. The AI transcription engine detects the language automatically and returns a speaker-labeled transcript.

2

Copy a prompt

Click the Copy button on any prompt below. The full prompt — including instructions and attribution — copies to your clipboard.

3

Paste & process

Open ChatGPT, Claude, or any AI tool. Paste the prompt, then paste your transcript where indicated. Submit for structured output.

1

Translate Transcript to English

BrassTranscripts produces speaker-labeled transcripts in the source language. This prompt takes that output and produces a full English translation that preserves every [SPEAKER_01] label and timestamp — so the conversation structure stays intact.

Best for: Dutch, Spanish, French, German, and Portuguese audio recordings where the downstream audience reads English.

You are a professional translator. I will provide you with a transcript that may be in a non-English language or contain non-English passages. Your task is to produce a full, accurate English translation. Instructions: - Translate ALL spoken content into English - Preserve every speaker label exactly as written (e.g., [SPEAKER_01], [SPEAKER_02]) — do not translate or remove them - Preserve all timestamps in their original format (e.g., [00:01:23]) - If a word or phrase is a proper noun (name of a person, company, city, product), keep it in the original language and note it in parentheses only if the meaning would be unclear - Maintain the speaker's tone — formal stays formal, casual stays casual - If a section is inaudible or marked [INAUDIBLE], keep that label in place - For [TARGET_LANGUAGE: Dutch / Spanish / French / German / other], note the source language at the top of your output Output format: Source language detected: [LANGUAGE] --- [SPEAKER_01] [00:00:05]: [English translation here] [SPEAKER_02] [00:00:18]: [English translation here] [continue for all lines...] Transcript: [PASTE YOUR BRASSTRANSCRIPTS OUTPUT HERE] --- Prompt by BrassTranscripts (brasstranscripts.com) – Professional AI transcription with professional-grade accuracy. ---
2

Multilingual Meeting Summary

International team meetings often involve two or more languages in the same recording. This prompt reads the full transcript regardless of language and produces a single, structured English summary — decisions, action items, and speaker attribution included.

Best for: International team meetings, cross-border client calls, and multilingual board sessions where stakeholders need a single English record.

You are an expert meeting facilitator and multilingual analyst. I will provide you with a transcript from a meeting where participants spoke in one or more languages. Produce a single, unified English summary. Instructions: - Read the entire transcript regardless of language(s) present - Identify each speaker by their label ([SPEAKER_01], [SPEAKER_02], etc.) and note their apparent language or role if determinable from context - Produce the summary entirely in English, even if the meeting was in another language - Do not translate the full transcript — summarize only Output the following sections: ## Meeting Overview One paragraph (3–5 sentences) covering the meeting purpose, participants, and overall outcome. ## Languages Present List each language identified in the transcript and which speaker(s) used it. ## Key Discussion Points Bullet list of the main topics covered. For each topic, note which speaker(s) drove the discussion. ## Decisions Made Numbered list of concrete decisions reached during the meeting. ## Action Items Table with columns: Action | Owner (speaker label or role) | Deadline (if mentioned) ## Open Questions Any unresolved issues or questions raised but not answered. Transcript: [PASTE YOUR BRASSTRANSCRIPTS OUTPUT HERE] --- Prompt by BrassTranscripts (brasstranscripts.com) – Professional AI transcription with professional-grade accuracy. ---
3

Language Identification

Before translating or analyzing a multilingual transcript, you need to know exactly which languages are present and who uses them. This prompt scans the full transcript, identifies each language, maps it to the speaker labels, and reports confidence levels and first-occurrence timestamps.

Best for: Researchers, content archivists, and QA reviewers who need a documented record of language distribution before downstream processing.

You are a computational linguist. Analyze the following transcript and identify every language present. Instructions: - Scan the full transcript for spoken segments in any language other than English, or confirm if only one language is used throughout - For each language detected, identify which speaker label(s) used it and provide representative timestamps - If a segment is ambiguous (e.g., heavy accent but still English, or very short utterance), note it as "uncertain — likely [LANGUAGE]" - Do not translate content — identification only Output format: ## Language Identification Report ### Languages Detected | Language | Confidence | Speakers Using It | First Timestamp | |----------|------------|-------------------|-----------------| | [Language] | High / Medium / Low | [SPEAKER_0X] | [HH:MM:SS] | ### Speaker Language Breakdown For each speaker label found in the transcript: - [SPEAKER_01]: Primary language — [LANGUAGE]. Timestamps: [list key timestamps] - [SPEAKER_02]: Primary language — [LANGUAGE]. Code-switches to [LANGUAGE] at [timestamps] ### Monolingual / Multilingual Assessment State clearly: Is this transcript monolingual, bilingual, or multilingual? ### Notes Any observations about accent variation, dialect, or unusual language patterns. Transcript: [PASTE YOUR BRASSTRANSCRIPTS OUTPUT HERE] --- Prompt by BrassTranscripts (brasstranscripts.com) – Professional AI transcription with professional-grade accuracy. ---
4

Code-Switching Analysis

Code-switching — alternating between languages within a conversation — is common in bilingual communities, international negotiations, and academic fieldwork. This prompt identifies every transition point, classifies whether the switch happened mid-sentence or between sentences, and surfaces observable patterns such as topic-triggered or participant-triggered switching.

Best for: Linguistics researchers, sociolinguistics educators, and anthropologists studying language behavior in bilingual or multilingual communities.

You are a sociolinguistics researcher specializing in code-switching analysis. Analyze the following transcript and produce a complete timeline of every language switch. Definition: A code-switch occurs when a speaker transitions from one language to another within or between utterances. Instructions: - Identify every code-switch event in the transcript - Record the speaker label, timestamp, the language switched FROM, the language switched TO, and quote the exact phrase or sentence where the switch occurs (first 10 words maximum) - Count total code-switch events per speaker - Note patterns (e.g., "Speaker 2 consistently switches to Spanish when discussing prices") - Distinguish intra-sentential switches (mid-sentence) from inter-sentential switches (at sentence boundaries) Output format: ## Code-Switching Timeline | # | Timestamp | Speaker | From Language | To Language | Switch Type | Opening Phrase | |---|-----------|---------|---------------|-------------|-------------|----------------| | 1 | [HH:MM:SS] | [SPEAKER_0X] | English | Spanish | Inter-sentential | "Bueno, entonces..." | ## Summary Statistics - Total code-switch events: [N] - Events per speaker: [SPEAKER_01: N, SPEAKER_02: N, ...] - Most common language pair switched: [LANG A ↔ LANG B] - Dominant switch direction: [LANG A → LANG B] ([N] times) vs. [LANG B → LANG A] ([N] times) ## Pattern Observations 3–5 bullet points describing observable patterns (topic-triggered, participant-triggered, formality-triggered, etc.) Transcript: [PASTE YOUR BRASSTRANSCRIPTS OUTPUT HERE] --- Prompt by BrassTranscripts (brasstranscripts.com) – Professional AI transcription with professional-grade accuracy. ---
5

Cross-Language Terminology Extraction

Proper nouns, brand names, and technical terms appear in their original language regardless of the meeting language — and AI transcription engines sometimes misspell them when they appear in a non-primary language context. This prompt extracts every proper noun and technical term across all languages, groups them by category, and flags potential transcription errors for human review.

Best for: Business intelligence analysts, CRM enrichment teams, and sales operations professionals who need clean entity data from international call recordings.

You are a multilingual business intelligence analyst. Extract all proper nouns, technical terms, and company-specific vocabulary from the following transcript, regardless of which language they appear in. Instructions: - Identify: personal names, organization names, product names, brand names, industry-specific terms, acronyms, place names, and recurring jargon - For each term, note: the term as it appears in the transcript, the language it was spoken in, any variant spellings or alternate forms heard, and a brief definition or context note if the meaning is unclear - Group by category Output format: ## Cross-Language Terminology Extraction Report ### People & Roles | Name / Label | Language Used | Context | |--------------|---------------|---------| | [Name] | [Language] | [Speaker label, role, or context] | ### Organizations & Brands | Term | Language | Notes | |------|----------|-------| | [Name] | [Language] | [Context or variant spelling] | ### Products & Services | Term | Language | Notes | |------|----------|-------| ### Technical & Industry Terms | Term | Language | Definition / Context | |------|----------|---------------------| ### Acronyms & Abbreviations | Acronym | Expanded Form | Language | Context | |---------|--------------|----------|---------| ### Variant Spellings / Ambiguous Terms List any terms where the AI transcription engine may have produced an incorrect spelling due to the speaker's accent or the term being in a non-primary language. Transcript: [PASTE YOUR BRASSTRANSCRIPTS OUTPUT HERE] --- Prompt by BrassTranscripts (brasstranscripts.com) – Professional AI transcription with professional-grade accuracy. ---
6

Multilingual Executive Brief

When leadership needs a concise English summary of a non-English meeting, a full translation is often too long to be useful. This prompt produces a tight 400-word brief — situation, decisions, action items, risks, and recommended follow-up — regardless of the source language or languages of the original recording.

Best for: Executives reviewing non-English subsidiary meetings, project managers receiving international client call summaries, and global operations teams that run meetings in the local language but report up in English.

You are an executive communications specialist. I will provide a transcript of a meeting conducted in one or more non-English languages. Produce a concise English executive brief suitable for a C-level reader who was not present. Requirements: - Maximum 400 words total - No direct quotations longer than one sentence - Use precise business language — no filler, no hedging - If the meeting language was not English, note the source language(s) at the top Output format: **Meeting Language(s):** [e.g., Dutch / Spanish / Mixed French-English] **Meeting Type:** [e.g., Weekly ops review / Client negotiation / Strategy session] **Date / Duration:** [from transcript metadata if available] --- ### Situation One paragraph: what was the context and why did this meeting take place? ### Key Decisions Numbered list — maximum 5 items. Each item: decision + the speaker/role who confirmed it. ### Action Items | Action | Owner | Deadline | |--------|-------|----------| | [Action] | [SPEAKER_0X or role] | [Date or "TBD"] | ### Risks & Blockers Bullet list of any risks, unresolved disagreements, or blockers mentioned. If none, write "None identified." ### Recommended Follow-Up 1–3 bullet points on what leadership should do next based on this meeting. Transcript: [PASTE YOUR BRASSTRANSCRIPTS OUTPUT HERE] --- Prompt by BrassTranscripts (brasstranscripts.com) – Professional AI transcription with professional-grade accuracy. ---
7

Multilingual Transcript Cleanup

AI transcription engines perform well on their primary training languages but introduce specific error types on accented or mixed-language audio: phonetic proper-noun misspellings, missing diacritical marks, false cognates, and currency or unit errors. This prompt targets those error types specifically — it does not rewrite or paraphrase, it corrects and logs.

Best for: Editors and QA reviewers cleaning up transcripts from non-native English speakers, bilingual interviews, and multilingual conference calls before the transcript is distributed or archived.

You are a professional transcript editor specializing in multilingual audio. The following transcript was produced by an AI transcription engine from audio that contained non-English speech, accented English, or mixed-language content. Identify and correct transcription errors specific to that context. Types of errors to correct: 1. **Proper noun misspellings** — names of people, companies, products, or places spelled phonetically by the AI instead of correctly 2. **False cognate errors** — words from another language that sound like English words but were transcribed as the English word (e.g., "actual" transcribed when speaker said "aktuell" in German) 3. **Homophone errors caused by accent** — English words misheard due to the speaker's native-language accent (e.g., "sheet" vs. "ship", "tree" vs. "three" in certain accents) 4. **Missing diacritics / accented characters** — proper nouns that should include é, ü, ñ, ç, etc. 5. **Numeric / unit errors** — currency amounts or measurements that don't match the language context (e.g., "$" when speaker clearly said "euros") Instructions: - Do NOT rewrite or paraphrase — correct only the specific error types above - For each correction, use the format: [ORIGINAL] → [CORRECTED] with a brief note - If a term is uncertain, flag it with [UNCERTAIN: possible correction] - Preserve all speaker labels, timestamps, and formatting exactly Output: ## Correction Log | # | Speaker | Timestamp | Original Text | Corrected Text | Error Type | Note | |---|---------|-----------|---------------|----------------|------------|------| ## Corrected Transcript [Full transcript with all corrections applied — same format as original] Transcript to clean: [PASTE YOUR BRASSTRANSCRIPTS OUTPUT HERE] --- Prompt by BrassTranscripts (brasstranscripts.com) – Professional AI transcription with professional-grade accuracy. ---
8

Subtitle Localization Prep

Localization projects need transcripts segmented by speaker, timestamp, and language — not just a flat text file. This prompt converts any BrassTranscripts output into a structured segment table tagged by language and code-switch status, ready to hand off to a localization platform or subtitle editor without manual re-segmentation.

Best for: Video producers, subtitle editors, localization project managers, and e-learning developers working with multilingual source recordings that will be dubbed or subtitled in additional languages.

You are a localization engineer preparing transcripts for subtitle production. Process the following transcript and produce a structured localization-ready output segmented by speaker, timestamp, and detected language. Instructions: - Parse every transcript line into individual segments - For each segment, identify: speaker label, start timestamp, end timestamp (estimate from next segment if not present), detected language, and spoken text - Flag any segment where the language differs from the dominant language of that speaker (likely code-switch) - Keep each segment short enough for subtitle use — if a single speaker turn exceeds [MAX_CHARS_PER_SUBTITLE: 84] characters, split it at a natural sentence boundary - Do not translate — this is a segmentation and language-tagging step only Output format: ## Localization Prep — Segment Table | Seg # | Speaker | Start | End | Language | Code-Switch? | Text | |-------|---------|-------|-----|----------|--------------|------| | 001 | [SPEAKER_01] | 00:00:04 | 00:00:09 | Dutch | No | [Text here] | | 002 | [SPEAKER_02] | 00:00:10 | 00:00:15 | English | Yes (from Dutch) | [Text here] | ## Language Distribution Summary | Language | Segment Count | % of Total | Speakers | |----------|--------------|------------|---------| ## Localization Flags List any segments that require special attention for localizers: - Segments with proper nouns that should NOT be translated - Segments where speaker emotion or tone affects subtitle timing - Segments marked [INAUDIBLE] or [UNCLEAR] ## Recommended Output Files Based on the language distribution, suggest which subtitle tracks to produce (e.g., "English translation track for Dutch source", "Spanish track for English segments"). Transcript: [PASTE YOUR BRASSTRANSCRIPTS OUTPUT HERE] --- Prompt by BrassTranscripts (brasstranscripts.com) – Professional AI transcription with professional-grade accuracy. ---

Frequently Asked Questions

Can BrassTranscripts handle non-English audio?

Yes. BrassTranscripts supports 102 languages via the FLEURS benchmark language set. Upload audio in Dutch, Spanish, French, German, Mandarin, Arabic, Portuguese, or dozens of other languages — the AI transcription engine detects the language automatically and produces a labeled transcript with no manual language selection required.

What is code-switching in a transcript?

Code-switching is when a speaker alternates between two or more languages within the same conversation. It is common in bilingual households, international business meetings, and academic interviews. The Code-Switching Analysis prompt on this page identifies every transition point and outputs a timestamped language-switch timeline organized by speaker.

Do these prompts work with ChatGPT and Claude?

Yes. All 8 prompts on this page are designed for any AI chat tool — ChatGPT (GPT-4o), Claude, Gemini, Perplexity, or any LLM that accepts long text input. Copy the prompt, paste your BrassTranscripts output below it, and submit. No special configuration is required.

How do I translate a transcript to English using AI?

Use the Translate Transcript to English prompt at the top of this page. Paste your non-English BrassTranscripts output at the bottom of the prompt and submit it to ChatGPT or Claude. The AI will produce a full English translation that preserves the original [SPEAKER_01], [SPEAKER_02] labels so you can still follow the conversation by participant.

What languages work best with these multilingual prompts?

The prompts work with any language supported by large language models. In practice, the best results come from European languages (Spanish, French, German, Dutch, Portuguese, Italian) and East Asian languages (Mandarin, Japanese, Korean). Arabic, Hindi, and Turkish also perform well. For lower-resource languages, the AI may produce less fluent translations — verify against the original when precision matters.

Can I download these prompts as a file?

Yes. The complete prompt library — including all multilingual prompts — is available on GitHub at github.com/CopperSunDev/brasstranscripts-ai-prompts in Markdown and YAML format. You can clone the repository or download individual prompt files.

Ready to transcribe multilingual audio?

Upload your audio in any of 102 supported languages. The AI transcription engine detects the language automatically and returns a speaker-labeled transcript — ready to use with any prompt on this page.

Upload Your Audio File

Need the full prompt library? Browse all 130+ prompts