TTS Swap › Guides

From Swift Labs, the team behind TTS Swap. This guide was written with AI help and checked against our own service. Not affiliated with OpenAI.

What is happening

OpenAI's deprecations page lists tts-1, tts-1-hd and gpt-4o-mini-tts (including the -2025-03-20 and -2025-12-15 snapshots) as shutting down on January 6, 2027. The listed replacement is gpt-realtime-2.1-mini. After that date, requests to /v1/audio/speech with those models stop working. Our migration guide walks through the options with curl, Node and Python snippets.

You have three broad choices:

1. Move to OpenAI's replacement. Best if you want to stay on OpenAI voices. Expect to rework code around the new model.
2. Move to another hosted TTS provider. Usually a new SDK and request shape.
3. Keep your code and change the endpoint. That is what TTS Swap is for. See the tts-1 migration page for a drop-in snippet.

This guide covers option 3, and is honest about the tradeoffs so you can decide if it fits.

Step 1: Find your speech calls

Search your code for audio.speech.create (Python or Node) or raw POSTs to /v1/audio/speech. Note which response_format, voice and speed values you use; they matter in step 4.

Step 2: Get a key

Buy a key at https://tts.deadlinkpatrol.com/ for $19/month. The key shows on screen right after checkout. It includes 1,000,000 characters per month. You can cancel yourself from the Stripe billing page linked on the site.

Or start with a free key: 10,000 characters a month, 5 requests a minute, one per email. Just an email, no card.

Step 3: Change your base URL and API key

Python:

from openai import OpenAI
client = OpenAI(
    base_url="https://tts.deadlinkpatrol.com/v1",
    api_key="sk-tts-…",
)
audio = client.audio.speech.create(model="tts-1", voice="alloy", input="Hello")
audio.write_to_file("hello.wav")

Node works the same way: pass baseURL and apiKey to the OpenAI constructor. Your audio.speech.create calls stay the same. More copy-paste snippets: curl, Node, Python.

Step 4: Check the differences

Read these before you switch production traffic.

• The voice is different. Audio comes from the open-source MeloTTS model on Cloudflare Workers AI, with one voice per language. Every voice name (alloy, nova, onyx and the rest) is accepted and gets that same voice. Listen to the English, Spanish and French samples on the home page first.
• Output is WAV. You get WAV even when you ask for mp3, the default. Browsers, ffmpeg and most players read it fine. If you feed the bytes to an MP3-only decoder, convert first, for example ffmpeg -i in.wav out.mp3. Save files with a .wav extension.
• pcm works like OpenAI: raw 24 kHz, 16-bit, mono.
• opus, aac and flac return a clear 400 error (unsupported_value). Switch those calls to wav or pcm.
• speed is accepted but not applied yet.
• Languages: English by default. For Spanish, French, Chinese, Japanese or Korean, add extra_body={"language":"es"} (or fr, zh, ja, ko).
• Input size: up to 4,096 characters per request, like OpenAI.

Step 5: Handle limits

Paid keys allow 20 requests per minute. There is also a shared daily capacity cap for the whole service, and no SLA. Over a limit you get an OpenAI-style 429 with a Retry-After header, and you are not charged characters for it. If your code already retries OpenAI 429s, it will retry these too.

To check your usage, call GET https://tts.deadlinkpatrol.com/v1/usage with your key. It returns the month, characters used and your limit. Each speech response also carries x-ttsswap-chars-used and x-ttsswap-chars-limit headers.

Step 6: Test, then switch

1. Point a staging environment at the new base URL.
2. Generate a few real inputs and listen to them.
3. Confirm your playback or storage path accepts WAV.
4. Switch production well before January 6, 2027, so you are not debugging on the day it breaks.

When TTS Swap is not the right choice

• You need a specific OpenAI voice, or several distinct voices per language.
• You need MP3 or Opus bytes and can't convert.
• You need adjustable speed today, or a guaranteed SLA.

In those cases, OpenAI's replacement or another provider is a better fit.

Privacy

Our privacy page says text and audio are not stored; we keep a hash of your key, your email and IDs. See /terms and /privacy on the site for details.

Questions or billing: swiftlabs-ops@agentmail.to.

FAQ

When does tts-1 stop working? OpenAI's deprecations page lists January 6, 2027 for tts-1, tts-1-hd and gpt-4o-mini-tts.

How much does TTS Swap cost? $19/month for 1,000,000 characters. Or try a free key first (10,000 characters a month). Cancel yourself from the Stripe billing page linked on the site.

Is it the same voice? No. It's a different, open-source voice (MeloTTS), and audio comes back as WAV. Hear the samples on the home page before you pay.

Related: What changes when you point tts-1 code at TTS Swap · Migration guide

TTS Swap by Swift Labs. Written with AI help. Not affiliated with OpenAI. · Terms · Privacy · Manage or cancel subscription