feat(OpenAI Node): Add gpt-4o-mini-tts model with new voices and parameters

[everettbu]

Mirror of n8n-io/n8n#21944
Original author: mo-sil

---

Summary

Adds complete support for OpenAI's latest Text-to-Speech features, bringing the node to 100% API parity with the OpenAI Audio API.

Changes

New Model

• ✅ gpt-4o-mini-tts - OpenAI's next-generation audio model with improved steerability

New Voices (5)

• ✅ ash - Available on all models
• ✅ coral - Available on all models
• ✅ sage - Available on all models
• ✅ ballad - gpt-4o-mini-tts exclusive
• ✅ verse - gpt-4o-mini-tts exclusive

New Audio Formats (2)

• ✅ wav - Waveform Audio File Format
• ✅ pcm - Raw 16-bit low-endian PCM audio

New Parameters

• ✅ instructions - Control voice characteristics (gpt-4o-mini-tts only)
• ✅ stream_format - Choose between audio or sse streaming

Implementation Details

• Conditional voice display: ballad and verse only appear when gpt-4o-mini-tts is selected
• DRY implementation: Voice lists are defined once and reused
• Voices automatically sorted alphabetically for better UX

Testing

• ✅ 37 comprehensive test cases added
• ✅ Tests cover all models, voices, formats, and parameters
• ✅ Separate test suites for common and model-exclusive voices
• ✅ All tests passing

Validation

• ✅ Tested with actual OpenAI API
• ✅ Verified voice exclusivity (ballad/verse only work with gpt-4o-mini-tts)
• ✅ All linting checks pass
• ✅ TypeScript compliant

Breaking Changes

None. This is a backward-compatible feature addition.

Related

• OpenAI API Reference: https://platform.openai.com/docs/api-reference/audio/createSpeech
• Models with enhanced steerability: https://openai.com/index/introducing-our-next-generation-audio-models/

---

Checklist:

• Code follows n8n style guide
• No ts-ignore used
• Tests included and passing
• No repetitive code (DRY principle applied)
• Single feature focus (TTS enhancements)
• TypeScript compliant

---

Note

Adds gpt-4o-mini-tts support with expanded voice options, new audio formats, instructions/stream_format params, and comprehensive tests for audio generation.

• OpenAI Audio TTS (packages/@n8n/nodes-langchain/nodes/vendors/OpenAi/v2/actions/audio/generate.operation.ts):
  • Model: Add gpt-4o-mini-tts.
  • Voices:
    • Split voice options by model; reuse shared list commonVoices and add model-specific ballad/verse for gpt-4o-mini-tts (alphabetically sorted).
  • Formats: Add wav and pcm to response_format options.
  • Params: Add instructions and stream_format (exposed only for gpt-4o-mini-tts) and include in request body when provided.
  • Behavior: Streamed request via POST /audio/speech; supports custom binary output field.
• Tests (packages/@n8n/nodes-langchain/nodes/vendors/OpenAi/test/v2/actions/audio/generate.operation.test.ts):
  • Add extensive tests covering models, voices (common and exclusive), formats, speed, instructions, stream_format, and custom binary field.

^{Written by Cursor Bugbot for commit 94f9c34. This will update automatically on new commits. Configure here.}