Free Text-to-Speech API Credits: Give Your Companion a Voice (TTS)
Your companion's voice is half her personality. Picking a text-to-speech (TTS) provider is really about languages and voices — the biggest credit pile means nothing if there's no voice you like in the language you speak. And this link in the chain has a perk the others don't: Edge TTS is completely free, with no registration and no API key, so you can hear her talk before spending a cent.
Four things to weigh before picking a TTS provider
1. Languages
Confirm the provider reads your language well. Chinese is everywhere, but Japanese — a must for many anime companions — isn't equally good everywhere. Language lists grow with updates, so after registering, read a paragraph in each language you plan to use.
2. Voices — this is the real decision
Two voices from the same provider can feel worlds apart. Audition on the provider's voice page before you register and confirm there's a Chinese or Japanese voice you actually like: some providers win on expressiveness (MiMo's Bingtang, MiniMax's character voices), others on marketplace scale (Fish Audio ships community-made anime voices and clones a custom voice from 10–15 seconds of audio).
3. Latency
Speech is the last hop of the conversation, and latency decides how quickly she picks up her line. Domestic nodes (MiMo, Qwen, MiniMax for users in China) respond fastest; overseas providers (ElevenLabs, Fish Audio) depend on your route. AniMate splits text into chunks and requests them in segments, so TTS is slightly more forgiving than STT — the node logic is the same.
4. Cost
TTS bills by character count. Paths to zero: start with Edge TTS; use limited-time free tiers (MiMo's TTS models are officially free for now); spend referral credits (MiMo ¥10 for both sides, SiliconFlow's ¥16 voucher); and remember a few thousand characters a day is typical, so character-priced free tiers vary a lot in how long they last.
Free tiers, languages, and voices at a glance
Compiled August 2026, covering every TTS provider AniMate supports. Voices and promotions shift — audition before you commit.
| Provider | Languages | Chinese / Japanese voices | Free tier |
|---|---|---|---|
| Edge TTS | Chinese (Mandarin), English, Japanese, and dozens more | Zh: Xiaoxiao, Yunjian, Yunze; Ja: Nanami, Keita | Free forever, no key, no sign-up |
| Xiaomi MiMo TTS | Chinese, English | Bingtang, Baihua, Moli, Soda — 2026 models with strong expressiveness | Free for a limited time (no end date announced) |
| Qwen AI (Alibaba) | Qwen3-TTS: 11 languages; CosyVoice: Mandarin, Cantonese, Sichuanese and other dialects | Qwen3-TTS: 51 expressive voices; CosyVoice: dozens of system voices + voice cloning | 90-day new-user quota; 1,000 free voice-clone creations |
| MiniMax | 40 languages (Chinese, English, Japanese, Korean...) | 17+ preset personas (energetic girl, gentle executive...) | Sign-up credits (amount shown in console) |
| Fish Audio | 83 languages (Chinese, English, Japanese included) | Marketplace full of community voices; 10–15 second cloning | S2.1 Pro free API (fair use, limited-time) |
| ElevenLabs | 74 languages (Multilingual v2 includes Chinese and Japanese) | Reads Chinese/Japanese; voices via the Voice Library | 10,000 characters/month (~minutes a day) |
| SiliconFlow | CosyVoice2: Chinese-focused, 150ms latency | CosyVoice2 system voices, dependable quality | ¥14 sign-up credit (¥16 voucher after verification) |
| Deepgram Aura | English, Spanish, French, German, Dutch, Italian, Portuguese (7) | No Chinese/Japanese — best for English-speaking users | Shares the $200 credit pool |
| Speechmatics TTS | English only | English voices | Shares the $100 credit pool |
Start at $0: Edge TTS (no key, no sign-up)
Edge TTS rides Microsoft Edge's read-aloud endpoint, backed by Azure's voice catalog: Chinese has Xiaoxiao (long the benchmark for free Chinese TTS), Yunjian, Yunze; Japanese has Nanami and more. Pick the Edge preset in AniMate and she's talking — the shortest path to a full voice conversation.
- Great for: hearing the full voice-conversation experience at zero cost; everyday use without high demands.
- Know first: it's not a supported commercial API — Microsoft can change the server side and break it until the community catches up, with no SLA. If that bothers you, use a provider's free tier below.
Provider-by-provider reviews
Xiaomi MiMo TTS: free for now, with voice design built in
Free tier: The mimo-v2.5-tts models are officially free for a limited time (no end date announced); the platform account and API key are shared with MiMo's LLM, and the ¥10 trial credit covers both.
Languages and voices: Chinese and English; preset voices Bingtang, Baihua, Moli, and Soda each have one-click previews, plus text-described voice design (voicedesign) and cloning (voiceclone).
Heads-up: Watch the official announcements for the free-period end date; if you already registered through the invite link on the LLM guide, just reuse the account.
Qwen AI (Alibaba): the deepest Chinese-dialect voice bench
Free tier: 90-day new-user quota; CosyVoice voice cloning includes 1,000 free creations for new users.
Languages and voices: Qwen3-TTS-Flash offers 51 expressive voices across 11 languages; CosyVoice's dozens of system voices cover Mandarin, Cantonese, Sichuanese, and more — the only provider that can give your companion a dialect.
Heads-up: Create a key at platform.qianwenai.com; enable the spend cap, and confirm quota terms in the console.
MiniMax: the widest preset voice shelf
Free tier: Sign-up credits (amount shown in your console); the speech-2.8 series bills per character.
Languages and voices: 40 languages including Chinese, English, Japanese, and Korean; 17+ preset personas you can preview in the official voice library.
Heads-up: The account is shared with MiniMax's LLM; Japanese is supported, but confirm exact preset voices in the library.
Fish Audio: voice marketplace and fast cloning
Free tier: The S2.1 Pro free API under fair use (a limited-time offer) spans 83 languages.
Languages and voices: The draw is voices — the marketplace carries community-made anime-style voices in Chinese, Japanese, and English, and 10–15 seconds of authorized audio is enough to clone your own.
Heads-up: Free-tier terms change quickly; check the site's current free-API page. Only clone voices you have permission to use.
ElevenLabs: broad languages, small free tier
Free tier: 10,000 characters monthly — with a few thousand characters of daily reading, that's a taste, not a workhorse.
Languages and voices: Multilingual v2 reads 74 languages including Chinese and Japanese, with a well-regarded Voice Library.
Heads-up: If you also use ElevenLabs for speech-to-text, one account covers both sides of the conversation.
SiliconFlow (via OpenAI Compatible): low-latency CosyVoice2
Free tier: ¥14 sign-up credit works for TTS; a ¥16 voucher after identity verification. CosyVoice2's limited-time free period has ended and it now bills per use.
Languages and voices: CosyVoice2 is Chinese-focused with 150ms latency and dependable system voices.
Heads-up: Connect through the OpenAI Compatible TTS preset; if you already registered via the LLM guide, reuse the same account and key.
Deepgram / Speechmatics TTS: the English-speaking extras
Both are supported TTS providers in AniMate and share their credit pools with the matching STT accounts: Deepgram Aura reads English, Spanish, French, German, Dutch, Italian, and Portuguese, while Speechmatics TTS is English only. If your companion speaks Chinese or Japanese, the providers above are the better fit — these two round out English-language characters.
Plug the key back into AniMate
- Open Settings → AI Services in AniMate and click "Add TTS" under speech services.
- Pick the provider and paste your API key — the Edge preset needs no credentials at all.
- Add a model and bind voices: fetch from the provider's voice library or enter voice IDs manually.
- Save, hit "Preview" in the chat, and set it as default — your companion has her voice.
FAQ
How many characters does daily speech use? Will free tiers last?
A few thousand characters a day is typical. Edge TTS and MiMo's limited-time free tier carry no quota pressure; character-priced tiers around 10,000/month (like ElevenLabs) are only enough to taste. For a daily driver, use a big credit pile or a limited-time free model.
Will Edge TTS stay free?
It's a browser read-aloud endpoint, not a commercial API, and Microsoft makes no stability promises. If it breaks, wait for the community library to update — or switch to a provider's free credits in the meantime.
How do I get a more anime-style voice?
Two routes: browse Fish Audio's marketplace for community-made anime voices, or clone a custom voice from 10–15 seconds of authorized audio with Fish Audio or CosyVoice.
Can I configure multiple TTS providers at once?
Yes — multiple TTS connections, each with its own models and voices, plus per-character voice bindings.
All three pieces — start talking
Chat model, ears, voice: the free credits for all three are in these guides. Once they're in, your companion can hear you and answer out loud.