An audio project should be a conversation, not a chain of disconnected tools
Producing useful audio usually involves more than generating a voice. A creator may need to read a document, transcribe a reference recording, divide a script into scenes, choose a speaking style, add music, review the result, and revise one section without losing the rest of the project.
The Miraga Audio Creation Agent brings those steps into one conversational workspace. Describe the outcome in ordinary language, attach the material you already have, and continue refining the same project until the audio is ready to use.
What you can start with
- A short idea: describe the subject, audience, language, duration, and tone.
- A finished script: paste the text and ask for a voice-over or speech recording.
- A document: upload PDF, TXT, or Markdown material for the Agent to read and adapt.
- A reference recording: upload audio for transcription, translation, pacing analysis, or creative direction.
- An existing project: return to the conversation and request a revised voice, wording, language, or scene.
One workspace, three useful views
The project list on the left keeps separate jobs easy to find and rename. The conversation in the center records the brief, progress, delivery notes, and every revision request. The panel on the right keeps inputs, scenes, scene audio, and the final assembled file together.
Long scene descriptions stay compact until you open the detailed view. When audio is available, it can be played directly from the corresponding scene or from the final delivery section.
How the Agent works through a request
- Understand the material: read the prompt and inspect attached text, documents, or audio.
- Plan the production: decide how many scenes are needed and define their purpose, dialogue, timing, voice, music, and transitions.
- Prepare the audio: check that the account can generate, create each required scene, and record the result in the project.
- Review and assemble: validate spoken content, duration, and endings, then assemble approved material into a final file.
- Continue the conversation: revise only what needs to change instead of restarting the entire project.
Built for common production scenarios
Voice-over
Turn product copy, tutorials, announcements, or video scripts into polished narration. Specify the intended audience and ask for a warm, authoritative, energetic, calm, or conversational delivery.
Audio translation
Upload a recording or provide a script, choose the target language, and request a localized version. The project can preserve meaning while adapting phrasing so it sounds natural when spoken.
Podcasts
Transform notes into a structured episode with an opening, sections, speaker turns, transitions, and a closing. Scenes make longer productions easier to review and replace.
Speeches and presentations
Create an audio version of a keynote, lesson, training module, or internal presentation. The Agent can help organize pacing, pauses, emphasis, and subtle background music.
Multilingual by design
The Agent follows the language used in the conversation unless you request another one. Audio projects can be prepared in English, Chinese, Traditional Chinese, Japanese, and Korean, including projects where the source and delivery languages are different.
For names, brands, abbreviations, and technical terms, include a pronunciation note. A short glossary is often the simplest way to keep several language versions consistent.
Files, privacy, and project continuity
Uploaded and generated files are stored inside a user- and project-specific location. Upload, playback, and download use temporary access, so project files do not need to be public. Each conversation keeps its own scenes, files, progress, and delivery history.
Projects can be renamed for easier retrieval and deleted when they are no longer needed. Long-running work can reconnect and recover the latest progress instead of depending on one uninterrupted browser request.
Credits are tied to successful generation
The Agent checks the available balance before generating audio. Transcription does not consume generation credits. For generated audio, the completed duration is used for billing, while failed generation attempts are not charged as successful audio.
A simple first project
Create a 20-second product welcome voice-over in English.
Use one warm, professional female speaker.
Keep the pace natural, add subtle music, and leave a clean music tail after the final sentence.
Deliver one playable final MP3.You do not need to name internal tools or describe the technical process. If the first version is close but not perfect, continue with a direct revision such as “make the voice calmer,” “shorten the music intro,” or “replace the last sentence without changing the rest.”
Where to go next
For a repeatable production method, continue with the Audio Creation Agent best-practices guide. If your project is centered on localization, see the voice-over and audio translation workflow. For longer spoken formats, read the podcast and speech production guide.