tts-1 is OpenAI's text-to-speech model optimized for real-time use cases, converting prepared text into playable speech. It is suitable for application prompts, content narration, and announcements in voice interactions, offering six preset voices as well as audio format and speech speed options. Compared with tts-1-hd, which prioritizes audio quality, tts-1 is better suited for tasks that prioritize responsiveness.
Submit requests to the public API at api.acedata.cloud using the documented parameters, then use the results in your application.
Input parameters and result formats vary by service. Use the public API for this model and follow its guide for generation, task retrieval and editing operations.
Clarify capacity, input and output, and invocation methods before selecting a model.
Real-time application positioning and preset voices are publicly available native capabilities; use formats, speech speed, and request methods through this platform's speech generation endpoint.
Learn what tts-1 can bring to your work.
The core trade-off of tts-1 is its focus on real-time use rather than making refined audio quality the sole objective. For action feedback, prompts, and narrated conversational responses, you can first prepare concise text, then generate speech and connect it to a player. This positioning is suitable for applications that value responsiveness, but it does not imply a fixed generation time.
Six preset voices allow applications to audition the same text and select a voice that suits the content style, then maintain consistency in subsequent announcements. voice uses fixed names and does not require uploaded recordings. When selecting a model, compare how real copy sounds rather than inferring gender, accent, or emotion solely from voice names.
When generating speech, you can choose the audio format and adjust the playback pace through speed. The default mp3 is convenient for integration with common players, while other formats can be selected according to the delivery pipeline. The returned content is audio data, and applications should save or play it in the selected format rather than parse the result as chat text or a download URL.
Start with specific tasks to find where the model can be effective.
Organize order statuses, operation steps, or device prompts into clear short sentences, select a fixed voice to generate spoken audio, and deliver it to the application player. For prompts used repeatedly, generated audio can be saved on the business side; for dynamic information, first complete text validation, then generate speech, to avoid reading incorrect content directly to users.
Enter edited article paragraphs, course explanations, or knowledge cards to output playable narration audio. It is recommended to organize content by semantic paragraphs, first listen to a sample containing terminology, numbers, and abbreviations, and then determine the voice and speaking rate. The deliverable is speech corresponding to the text; content rewriting and fact checking should be completed before generation.
In a voice assistant, tts-1 can handle the final step of converting response text to speech: first obtain a reply through business logic or a dialogue model, then submit text suitable for spoken expression for generation. It is responsible for speaking predetermined content and does not handle recording recognition or response reasoning, making it suitable for building voice interaction flows with clear responsibilities.
Choose based on task complexity, input materials, and expected results.
If the task involves frequent short prompts, immediate feedback, or dialogue playback, and response experience is the priority, tts-1 is a better fit. During evaluation, use real text, listen on target devices, and measure the complete playback chain; native real-time optimization does not mean that network, queuing, and playback startup times in an application are guaranteed.
tts-1-hd is a separate model focused on audio quality, not a voice option or output format for tts-1. For repeatedly played narration and polished content, compare both models using the same script, then decide based on listening quality and response requirements. Choosing a more refined format does not mean switching to the HD model; voice, format, and model should be considered separately.
Prepare the input first, then connect it to the corresponding application flow.
Prepare short prompt copy, the expected playback duration, and the selected voice, and standardize pronunciations for numbers and proper nouns.
Submit model=tts-1, input, and voice using /v1/audio/speech. Save the binary result according to the returned audio format, first listen in the actual player, then incorporate the chosen voice and format into content production.
Design tasks directly from the inputs and acceptance priorities below.
Submit model=tts-1, input, and voice via /v1/audio/speech; first preview with the default speed and MP3, then adjust the speaking rate.
Check response time, prompt intelligibility, and completeness at the beginning and end of sentences; in the real app, confirm that the audio format is playable and that text prompts appear on failure.
Before production use, understand the output quality and capability scope.
Answers to common questions about using tts-1.
The main difference is their optimization focus: tts-1 is designed for real-time use cases, while tts-1-hd prioritizes audio quality. For instant prompts and conversational announcements, consider tts-1 first; for polished narration, compare audio samples using the same script, as you cannot determine which model was used based on the file format alone.
You can choose alloy, echo, fable, onyx, nova, or shimmer, with alloy as the default voice. It is recommended to test each one with actual business scripts before settling on a voice. The names themselves do not promise an accent, emotion, or character, nor do they indicate the ability to replicate any person's voice.
Submit the input to be spoken to POST /v1/audio/speech, and explicitly set model to tts-1; add voice, response_format, and speed as needed. Explicitly selecting the model makes the request intent clearer, and the returned result should be handled as binary audio data.
Audio format options include mp3, opus, aac, flac, wav, and pcm, with mp3 as the default; speed defaults to 1.0. Choose a format based on your player and audio processing workflow, and listen after adjusting the speed rather than treating unspecified numeric ranges as usable limits.
You cannot treat tts-1 as a complete voice conversation model. It receives text and generates speech; if user input is a recording, it must first be transcribed, then response text generated, and finally passed to tts-1 for narration. This also makes it easier to review the content before playback and improve the spoken expression of the text.
Choose a development approach or application suited to this model and follow its setup guide.
↗Python SDK Integration Guide
SDK installation, authentication and examples↗TypeScript SDK Integration Guide
SDK installation, authentication and examples
↗Pricing
$12.74· per 1M characters
You pay$13,584for 142,800 credits · $0.0951 / credits46% OFF
You're billed in credits. Pick a top-up tier to see its per-credit rate — the larger the top-up, the cheaper each credit.
This is only a basic call example. See the full docs for more parameters and advanced usage.
View Docs