Workflow Recipes

Workflow Recipes are pre-designed step combinations for common tasks. Each recipe below lists the exact steps to chain together, plus key configuration tips to get the best results.

NOTE

In DuRT, open the Workflow panel, add steps in the listed order, configure each step, then press Run to execute the full pipeline.

---

Recipe 1: Quick File Transcription → Export SRT

Use case: You have a local audio or video file and want a subtitle file as fast as possible.

Steps

1. MediaTranscription
2. ExportArtifact

Configuration Tips

• MediaTranscription → Source: Select your local file (MP3, WAV, MP4, MKV, etc.).
• MediaTranscription → Word-level Timestamps: ✅ Enable — required for accurate SRT timing.
• MediaTranscription → Language: Set the spoken language explicitly rather than using auto-detect for best accuracy.
• ExportArtifact → Format: Select SRT.

TIP

If your video has multiple speakers, also enable Speaker Diarization to see who said what in the transcript.

---

Recipe 2: Meeting Recording → AI Summary → Export TXT

Use case: You recorded a meeting and want both a clean transcript and a concise summary, saved as plain text.

Steps

1. MediaTranscription
2. LLMProcess (Proofreading)
3. LLMProcess (Summary)
4. ExportArtifact

Configuration Tips

• MediaTranscription → Speaker Diarization: ✅ Enable to attribute text to each participant.
• LLMProcess #1 → Mode: Set to Proofreading to fix ASR errors and punctuation before summarising.
• LLMProcess #2 → Mode: Set to Summary.
• LLMProcess #2 → Custom Prompt: Optionally add instructions like "Focus on action items and decisions."
• ExportArtifact → Format: Select TXT.

TIP

Export both the proofread transcript and the summary in one step by selecting TXT and enabling separate export sections if your version supports it.

---

Recipe 3: Podcast Transcription → Translate to Chinese → Export Bilingual SRT

Use case: You want to watch or share an English-language podcast with Chinese subtitles, or produce a bilingual subtitle file.

Steps

1. MediaTranscription
2. LLMProcess (Translation)
3. ExportArtifact

Configuration Tips

• MediaTranscription → Source: Paste the YouTube/podcast URL, or select a local file.
• MediaTranscription → Word-level Timestamps: ✅ Enable for precise subtitle timing.
• LLMProcess → Mode: Translation.
• LLMProcess → Target Language: Chinese (Simplified) or your preferred locale.
• LLMProcess → Custom Prompt: Add "Preserve subtitle line breaks. Keep names and proper nouns in their original form."
• ExportArtifact → Format: SRT (or VTT for web players).

NOTE

For a true bilingual SRT (original + translation on separate lines), add a custom prompt instructing the LLM to output both the original line and its translation on consecutive lines.

---

Recipe 4: Live Captions → Export Transcript

Use case: You want to display real-time captions during a live event, meeting, or class — and save the full transcript afterwards.

Steps

1. RealtimeASR
2. ExportArtifact

Configuration Tips

• RealtimeASR → Input Source: System Audio for online meetings (captures all speakers); Microphone for in-person events.
• RealtimeASR → AEC: ✅ Enable if you have speakers on and are using System Audio input.
• RealtimeASR → Language: Set the spoken language.
• ExportArtifact → Format: TXT for a plain transcript, or SRT if you want timestamps.

IMPORTANT

Make sure macOS has granted DuRT Screen Recording permission (System Settings → Privacy & Security → Screen Recording) before using System Audio capture.

---

Recipe 5: Language Learning — Video Link → Transcription → Translation → TTS Playback

Use case: You are learning a language and want to transcribe a foreign-language video, translate it into your native language, and then listen to the translation read aloud for pronunciation practice.

Steps

1. MediaTranscription
2. LLMProcess (Translation)
3. TextToSpeech
4. ExportArtifact

Configuration Tips

• MediaTranscription → Source: Paste the video URL (YouTube, etc.) or select a local file.
• MediaTranscription → Word-level Timestamps: ✅ Enable.
• LLMProcess → Mode: Translation.
• LLMProcess → Target Language: Your native language.
• TextToSpeech → Voice: Choose a preset voice that matches the target language.
• TextToSpeech → Speed: Lower to 0.8× or 0.75× for easier comprehension while learning.
• ExportArtifact → Format: MP3 (audio) + SRT (subtitles) — export both for maximum flexibility.

TIP

After listening to the TTS audio, switch the voice to a different accent or gender to practise recognising the language in varied contexts.

---

Recipe 6: Video Dubbing Replacement — Transcription → Translation → TTS → Export Audio

Use case: You have a video in one language and want to produce a dubbed audio track in another language for re-mixing in a video editor.

Steps

1. MediaTranscription
2. LLMProcess (Proofreading)
3. LLMProcess (Translation)
4. TextToSpeech
5. ExportArtifact

Configuration Tips

• MediaTranscription → Source: Select the local video file.
• MediaTranscription → Word-level Timestamps: ✅ Enable — precise timing ensures the dub aligns with the original cuts.
• LLMProcess #1 → Mode: Proofreading — clean up any ASR errors before translation.
• LLMProcess #2 → Mode: Translation.
• LLMProcess #2 → Target Language: The dubbing language.
• LLMProcess #2 → Custom Prompt: "Adapt the translation for spoken dubbing. Keep sentence length natural and similar to the original."
• TextToSpeech → Voice: Use Voice Cloning if you want to match the original speaker's voice, or pick a professional preset.
• TextToSpeech → Speed: Adjust to approximately match the original speaking pace.
• ExportArtifact → Format: AAC or MP3 for the audio track; optionally also export SRT for the translated subtitle file.

TIP

Import the exported audio track into your video editor (Final Cut Pro, DaVinci Resolve, etc.) and mute the original audio track to complete the dub replacement.

NOTE

For Voice Cloning, provide a 10–30 second clean sample of the target voice before running the workflow.