

Alibaba's Qwen Audio 3.0 TTS family serves social, companion, and real-time applications. Guide Plus vocal delivery and volume with natural language instructions, or use Flash for live dialogue and customer service. Atlas Cloud offers standard pricing of $0.02 per 1,000 input characters for Plus and $0.015 for Flash, helping you estimate costs for each workload. Start building today.
Qwen Audio 3.0 TTS is developed by Alibaba. Atlas Cloud (operated by Atlas Cloud AI LLC) provides access to it and does not own it. All trademarks belong to their respective owners.
Atlas Cloud provides you with the latest industry-leading creative models.
Choose expressive flagship voices for companion experiences or low latency speech for real time interaction.
| Modality | Description |
|---|---|
| Qwen Audio 3.0 TTS Flash API (Text to Speech) | Turn text into speech with low latency synthesis, broad language support, and Chinese dialect coverage. Flexible instructions guide vocal delivery for voice assistants, live dialogue, and customer service. Standard pricing is $0.015 per 1,000 input characters. |
| Qwen Audio 3.0 TTS Plus API (Text to Speech) | For social and companion applications, this flagship endpoint converts text into natural, expressive speech using system voices. Control emotion, tone, role, speed, and volume through flexible instructions to shape each interaction. Standard pricing is $0.02 per 1,000 input characters. |
Choose Qwen Audio 3.0 TTS Flash or Plus for distinct conversational needs, with extensive voice catalogs, multilingual synthesis, adjustable delivery, structured text input, and flexible audio output through Atlas Cloud.
Qwen Audio 3.0 TTS offers Flash for low latency interaction and Plus for natural, expressive companion speech. Standard Atlas Cloud rates are $0.015 per 1,000 input characters for Flash and $0.02 for Plus. Choose the model that matches your product's conversational priorities, with pay-as-you-go billing. This gives voice assistants, customer service tools, and social applications two clearly priced synthesis options.
Each 3.0 model exposes a catalog of 597 voice-cloned base voices with preview samples and gender and language labels. Browse the available voices, listen to samples, and pass the selected model's voice ID with your text. Voice IDs are specific to their model. This makes casting a narrator or companion character easier before you integrate speech into your application.
For multilingual speech, both models accept one language hint per request, with 10 supported codes covering Chinese, English, Japanese, Korean, French, German, Portuguese, Italian, Vietnamese, and Indonesian. Set language_hints for the intended language and choose a compatible voice. Flash also supports Chinese dialect synthesis. Developers can use these controls to adapt spoken experiences for international audiences and regional Chinese communities.
Tell the model how a line should feel through natural language instructions for emotion, tone, persona, and delivery. Both models also expose speed and pitch multipliers from 0.5 to 2.0 and volume from 0 to 100. Combine a style instruction with these numeric controls to refine the same script. Expressive dialogue and companion experiences benefit from delivery tailored to the moment.
When a script needs deliberate pauses or spoken mathematics, both models support SSML and LaTeX input with enable_ssml enabled. Submit the prepared text through the same synthesis request rather than treating every passage as plain prose. SSML provides a way to mark pauses, while LaTeX represents formulas for speech. These input options suit educational narration and scripts that require more deliberate reading.
Audio delivery supports four formats: MP3, PCM, WAV, and Opus, with seven sample rate options from 8,000 to 48,000 Hz. Select output_format and sample_rate in your request; bit_rate applies only to Opus. Atlas Cloud uses asynchronous submission and task ID polling to retrieve generated audio URLs. This workflow lets teams prepare speech files for different playback and production requirements through one API.
Compare Qwen Audio 3.0 TTS with its flagship successor and an ElevenLabs alternative using identical scripts that invite subtle pauses, emotional shifts, and warm endings.
Shh... I thought we'd lost you. Wait, that light... you found the way home! Come here. You're safe now, and tonight, nobody has to face the dark alone.
Generated with Qwen-Audio 3.1 TTS Plus on Atlas Cloud
Generated with ElevenLabs v3 Text-to-Speech on Atlas Cloud
Generated with Qwen-Audio 3.0 TTS Plus on Atlas Cloud
Listen... just a little crackle. Now, the first bite: crisp crust, warm butter, pure joy! Oh, that's delicious. Pull up a chair. There's always room for you at our table.
Generated with Qwen-Audio 3.1 TTS Plus on Atlas Cloud
Generated with ElevenLabs v3 Text-to-Speech on Atlas Cloud
Generated with Qwen-Audio 3.0 TTS Plus on Atlas Cloud
Pair expressive Plus voices with latency-focused Flash synthesis in Qwen Audio 3.0 TTS to build companion dialogue, voice assistants, localized support, story narration, brand voiceovers, and spoken lessons.
Give virtual companions expressive voices with Plus, using instructions to shape emotion, tone, and role. Create warm greetings, playful exchanges, or reassuring replies for social apps and interactive character experiences.
When assistants need spoken replies, Flash provides speech synthesis tuned for low latency and live dialogue. Use it as the voice output layer for conversational interfaces that turn generated text into speech.
Flash supports multiple languages and Chinese dialects, bringing localized speech to customer service workflows. Turn written help responses into spoken guidance for regional support experiences tailored to the audiences your application serves.
With Plus, instructions shape a narrator's emotion and tone, while speech controls adjust pacing and volume. Produce spoken story passages and character lines for audiobook projects, storytelling apps, and serialized audio content.
Product scripts become expressive voiceovers through Plus, with instructions defining the delivery style and persona. Build spoken product introductions, promotional clips, and app walkthroughs around a voice direction that reflects your brand.
For learning applications, Plus converts instructional text into expressive speech with adjustable speed and volume. Educators can create narrated examples, lesson introductions, and spoken study guides with delivery suited to their teaching materials.
Compare Qwen Audio 3.0 TTS with the 3.1 variants and MiniMax Speech models by application focus, speech controls, audio formats and standard Atlas Cloud pricing.
| Model | Application Focus | Speech Controls | Audio Formats | Standard Price / 1K Characters |
|---|---|---|---|---|
| Qwen-Audio 3.0 TTS Plus | Expressive voices for social and companion experiences | Describe emotion, tone and role through instructions; adjust speed, pitch and volume | MP3, PCM, WAV, Opus | $0.02 |
| Qwen-Audio 3.0 TTS Flash | Low latency synthesis for assistants, live dialogue and customer service | Tune speed, pitch and volume alongside instructions for emotion and speaking style | MP3, PCM, WAV, Opus | $0.015 |
| Qwen-Audio 3.1 TTS Plus | Brand voices, audiobooks, dubbing and character performances | Direct emotional delivery and character roles with instructions; refine speed, pitch and volume | MP3, PCM, WAV, Opus | $0.03 |
| Qwen-Audio 3.1 TTS Flash | Multilingual and dialect speech for interactive applications | Combine natural language instructions with expressive tags and adjustable speech parameters | MP3, PCM, WAV, Opus | $0.003 |
| MiniMax Speech 2.8 HD | Speech production focused on naturalness and expressive quality | Select a voice and adjust speed, pitch and volume | MP3, WAV, FLAC | $0.10 |
| MiniMax Speech 2.8 Turbo | Fast speech generation with selectable voices and interjection tags | Use emotion presets and inline interjections; set speed, pitch and volume | MP3, WAV, FLAC | $0.06 |
Get started in minutes — follow these simple steps to integrate and deploy models through Atlas Cloud's platform.
Sign up at atlascloud.ai and complete verification. New users receive free credits to explore the platform and test models.
Combining the advanced Qwen Audio 3.0 TTS models with Atlas Cloud's GPU-accelerated platform provides unmatched performance, scalability, and developer experience.
Low Latency:
GPU-optimized inference for real-time reasoning.
Unified API:
Run Qwen Audio 3.0 TTS, GPT, Gemini, and DeepSeek with one integration.
Transparent Pricing:
Predictable per-token billing with serverless options.
Developer Experience:
SDKs, analytics, fine-tuning tools, and templates.
Reliability:
99.99% uptime, RBAC, and compliance-ready logging.
Security & Compliance:
SOC 2 Type II, HIPAA alignment, data sovereignty in US.
Qwen Audio 3.0 TTS is Alibaba's text-to-speech model family, available in Plus and Flash variants on Atlas Cloud. It converts written text into speech, with natural-language instructions for controlling emotion, tone, character, and delivery.
Build spoken experiences for social companions, voice assistants, interactive dialogue, and customer support. Expressive delivery makes Plus suitable for companion interactions, while Flash is tuned for applications that prioritize low-latency speech generation.
Create an Atlas Cloud API key, then send an authenticated POST request to https://api.atlascloud.ai/api/v1/model/generateAudio with model, text, and voice. Use alibaba/qwen-audio-3.0-tts-plus/text-to-speech or alibaba/qwen-audio-3.0-tts-flash/text-to-speech as the model ID. Poll https://api.atlascloud.ai/api/v1/model/prediction/{prediction_id} with the returned task ID until generation completes, then retrieve the audio URL from outputs.
Standard pricing on Atlas Cloud is $0.02 per 1,000 input characters for Plus and $0.015 per 1,000 input characters for Flash. Both variants use character-based, pay-as-you-go billing.
Each variant lists 597 base voices with language and gender labels and preview samples on its Atlas Cloud model page. The models support multilingual speech and Chinese dialects, with language hints including Chinese, English, Japanese, and Korean. Preview a voice in your target language or dialect before selecting it for production.
Describe the desired emotion, tone, or character in the instruction parameter, and adjust speed, pitch, and volume through their dedicated parameters. For structured text input, enable_ssml supports SSML and LaTeX, providing additional control over speech rendering.
Choose MP3, PCM, WAV, or Opus through output_format, with MP3 as the default. Both models also expose sample_rate, while bit_rate applies specifically to Opus output.
Choose Plus when natural, expressive delivery is central to a social or companion experience. If your application prioritizes low-latency interaction and lower character costs, Flash is the more suitable starting point. Compare both using the same script and desired speaking style before deciding.
The documented Atlas Cloud integration uses asynchronous submission and polling, returning an audio URL when the task completes. Treat this endpoint as a job-based workflow when designing playback, and confirm streaming availability separately if your application requires audio chunks during generation.
A voice ID belonging to another model can trigger InvalidParameter, so select a voice from the catalog associated with your chosen variant. When switching between Plus and Flash, update both model and voice rather than reusing the previous voice ID.
Guides, tutorials, and product updates to help you get the most out of Atlas Cloud.