Audio Node

Audio nodes are used to generate speech, sound effects, music, and other audio content, adding an auditory dimension to your creations.

Generation Modes

Audio nodes support multiple audio generation types:

⚠️ Important: Audio nodes do not accept connections from Text nodes. Text content is input directly within the Audio node.

Basic Usage

🗣️ Text to Speech

Convert text content into natural, fluent speech.

Steps:

1. Add an Audio node
2. Select Speech type at the top of the node
3. Select a TTS model (e.g., OpenAI TTS-1-HD, Minimax Speech 2.5)
4. Select a voice
5. Enter the text to be read in the input field
6. Click Generate

Text example:

Welcome to Martini, a powerful AI creative workflow platform.
Here, you can easily generate images, videos, and audio content using drag-and-drop nodes.

Use cases:

• Video narration
• Audiobooks
• Educational explanations
• Product introductions

🎵 Generate Sound Effects

Generate realistic sound effects based on descriptions.

Steps:

1. Add an Audio node
2. Select Sound Effects type
3. Select a model (recommended: ElevenLabs Sound Effects v2)
4. Enter sound effect description (English works better)
5. Set duration
6. Click Generate

Sound effect description examples:

Parameters:

• Duration: 0.5-22 seconds
• Prompt Influence: Degree of influence description has on result

🎼 Generate Music

Create background music or complete songs.

Steps:

1. Add an Audio node
2. Select Music type
3. Select a model (Suno V5 or Minimax Music v1.5)
4. Enter lyrics or style description
5. Set style tags (Genre, Mood)
6. Click Generate

Music description examples:

Suno advanced parameters:

• Style: Music style (Pop, Rock, Classical, etc.)
• Mood: Emotion (Happy, Sad, Energetic)
• Instrumental: Pure music (no vocals)
• Vocal Gender: Male / Female
• Weirdness: Creativity level

🎨 Voice Design

Create unique AI voices.

Steps:

1. Add an Audio node
2. Select Voice Design type
3. Select Minimax Voice Design model
4. Describe desired voice characteristics (gender, age, tone)
5. Enter preview text
6. Click Generate

Voice description example:

A young female voice, warm and friendly,
slightly husky, British accent

Uses:

• Create brand-specific voices
• Character voice design
• Diversified TTS voices

🎬 Video to Audio

Generate matching sound effects or background music based on video content.

Steps:

1. Prepare a Video node
2. Add an Audio node
3. Connect Video → Audio
4. Select Video to Audio type
5. Select Mirelo SFX V1 model
6. Set number of samples to generate (2-8)
7. Click Generate

Workflow:

Video → Audio (Mirelo SFX V1)

Features:

• AI analyzes video content
• Generates 2-8 different sound effect variants
• You can choose the most suitable one

Use cases:

• Add sound effects to silent videos
• Generate background music
• Create sound effect libraries

Model Selection Guide

(To be added)

Parameters

Voice

TTS models provide multiple preset voices.

OpenAI TTS-1-HD voices:

• Alloy: Neutral, balanced
• Echo: Male, steady
• Fable: British accent, narrative feel
• Onyx: Male, deep
• Nova: Female, energetic
• Shimmer: Female, gentle

Minimax Speech 2.5 voices:

• Supports multiple Chinese voices
• Supports emotion control (happy, sad, angry, etc.)

Speed

Adjust playback speed of speech (TTS mode).

Duration

Set the length of sound effects (Sound Effects mode).

ElevenLabs Sound Effects:

• Minimum: 0.5 seconds
• Maximum: 22 seconds
• Recommendation: Set based on actual needs

Emotion

Minimax Speech 2.5 supports emotion control.

Connection Rules

Audio nodes can receive from:

⚠️ Audio nodes do not accept Text node connections. Please enter text content directly within the Audio node.

Audio nodes can connect to:

Workflow Examples

🎙️ Video Audio

Video (silent video) → Audio (Mirelo) → Generate multiple sound effect options

🗣️ Digital Human Voice

Audio (TTS) ──┐
              ├→ Video (Kling AI Avatar)
Image (person) ─┘

First generate speech with Audio node, then connect to Video node to create digital human.

🎬 Complete Short Video Production

Text → Image → Video (silent video)

Generate separately:
- Audio (TTS) → Narration
- Audio (Sound Effects) → Background sound effects
- Audio (Music) → Background music

Finally composite in editing software

Upload Audio (as Starting Point)

You can upload local audio files to Audio nodes:

Method:

• Drag and drop audio file onto canvas
• Or click upload area within Audio node

Supported formats: MP3, WAV, M4A

Uses:

• Serve as input for Video nodes (create digital humans)
• Material for audio editing
• Export to other tools

Common Questions

❓ Why can't Audio nodes connect to Text nodes?

This is a design decision. Text input for Audio nodes is completed directly within the node, avoiding extra connection complexity.

Correct usage:

• ❌ Text → Audio (not supported)
• ✅ Enter text directly in Audio node

❓ How to choose the right TTS voice?

Recommended process:

1. First listen to previews of all voices
2. Choose based on content:
   • Serious content: Choose steady voices (Onyx, Echo)
   • Casual content: Choose lively voices (Nova, Shimmer)
   • Narrative content: Choose voices with storytelling feel (Fable, Ballad)
3. Can switch voices and regenerate if not satisfied

❓ Can music generation specify specific genres?

Yes! In Suno model:

Method 1: Use style tags

• Select from Style dropdown (Pop, Rock, Jazz, etc.)

Method 2: Describe in Prompt

80s synth-pop with retro drum machines,
nostalgic melody, upbeat tempo

❓ What if generated sound effect is too short/long?

Adjust Duration parameter:

• Sound Effects mode supports 0.5-22 seconds
• Directly adjust duration in parameter panel

If longer sound effects needed:

• Generate multiple sound effect segments
• Splice together in external audio editing software

❓ How to choose from multiple Video to Audio sound effects?

1. Mirelo generates 2-8 sound effect variants
2. Click audio waveforms within node to switch and preview
3. Select the most satisfactory one
4. Download or connect to downstream use

❓ Can generated speech speed be adjusted?

Yes! In TTS parameters:

• Find the Speed slider
• Adjust to 0.5x-2.0x
• Regenerate

Advanced Features

Minimax Emotion Control

Minimax Speech 2.5 supports fine-grained emotion and tone control.

Adjustable parameters:

• Emotion: Happy, Sad, Angry, Surprised, etc.
• Speed: Speaking rate
• Pitch: Tone
• Volume: Volume

Suitable for:

• Audiobooks (require rich emotional expression)
• Drama dubbing
• Advertisement videos

ElevenLabs Context Awareness

ElevenLabs TTS supports context input to improve naturalness.

Usage:

1. Enter previous text in Previous Text
2. Enter following text in Next Text
3. Current text will adjust tone based on context

Suitable for:

• Long-form reading (consistent tone between chapters)
• Dialogue scenes

Suno Custom Mode

Suno V5 supports highly customizable music generation.

Parameter control:

• Style Weight: Style intensity
• Weirdness Constraint: Creativity level
• Audio Weight: Melody weight

Suitable for:

• Experimental music
• Precise control of specific styles

Operation Buttons and Features

Generate

Click to start generating audio.

Generation time:

• TTS: 5-15 seconds
• Sound Effects: 10-30 seconds
• Music: 30-120 seconds

Play

Click play button to preview audio.

Features:

• Play/pause
• Volume adjustment
• Loop playback

Download

Download generated audio file.

Formats:

• TTS: MP3
• Sound Effects: WAV/MP3
• Music: MP3

Workflow Examples

🎙️ Create Audio Content

1. Audio (TTS) → Generate narration
2. Audio (Sound Effects) → Generate background sound effects
3. Audio (Music) → Generate background music

Mix and composite in external audio software (e.g., Audacity)

🎬 Complete Video Audio Workflow

Text → Image → Video (silent video)

Audio (TTS) → Narration audio
Audio (Video to Audio) ← Video → Generate ambient sound effects

Composite in editing software

🗣️ Digital Human Video

Audio (TTS, generate speech) ──┐
                       ├→ Video (Kling AI Avatar)
Image (person photo) ──────┘

See Video Node - Digital Human

Model Selection Guide

(To be added)

Common Usage Tips

📝 TTS Text Optimization

Punctuation affects pauses:

• Comma , = Brief pause
• Period . = Clear pause
• Question mark ? = Rising intonation
• Exclamation mark ! = Emphasis

Numbers and symbols:

• Write one hundred instead of 100 (unless you want "one zero zero")
• Write first instead of 1st

🎵 Sound Effect Generation Tips

Specific descriptions:

• ❌ water sound
• ✅ Heavy rain pouring on a tin roof

Add environment and distance:

• Close-up microphone of crackling fireplace
• Distant thunder rolling across hills

🎼 Music Generation Tips

Structured description:

1. Genre: Pop, Rock, Jazz, Classical
2. Instruments: Piano, Guitar, Synth, Orchestra
3. Mood: Upbeat, Melancholic, Epic
4. Tempo: Fast, Slow, Moderate tempo

Example:

Acoustic folk song with gentle guitar strumming,
warm male vocals, introspective lyrics,
slow tempo, indie style

Common Questions

❓ Which languages does TTS support?

❓ Are there copyright issues with generated music?

AI-generated music typically belongs to you for personal use, but it's recommended to:

• Check platform terms of use before commercial use
• Use original music for critical projects to be safe

❓ What if sound effect generation results are poor?

Optimization methods:

1. Use English descriptions: AI understands English more accurately
2. Add details: Describe sound effect texture, distance, environment
3. Adjust Duration: Sound effect length should be reasonable
4. Generate multiple times: Select best result

❓ How to add multiple audio layers to video?

Martini generates single-layer audio, multi-layer mixing requires external tools:

Recommended workflow:

1. Generate separately in Martini:
   • Narration (TTS)
   • Sound effects (Sound Effects)
   • Background music (Music)
2. Export all audio
3. Mix in Audacity / Premiere / Final Cut
4. Composite final audio with video

Performance Optimization Recommendations

(To be added)

Next Steps

• Video Node — Combine audio to create digital human videos
• Workflow Examples — Complete audio-video workflows