USAGE, ON YOUR TERMS
Free app.Pay only for what you use.
Choose who turns your speech into text, and who refines it. Use cloud services, local models, or both.
- No Vowrite subscription
- Switch providers independently
- Cloud or local refinement
A simple split: the app and your AI usage
- VOWRITE
- Free & open source
No Vowrite subscription to keep running.
- CLOUD AI
- Billed by usage
Paid directly to the providers you choose.
Local processing has no cloud API fee for that step.
YOUR PROVIDERS
Two steps. Your choice.
Recognition and refinement are independent. Switch either provider, refine with a local model, or skip refinement.
Speech to text
The recognition step.
Audio usage billed by your provider.
Text refinement
Optional: tidy or rephrase your words.
Billed for text processed (tokens).
Groq handles recognition in the cloud. DeepSeek bills input and output tokens separately.
This example changes only this page. Configure your real providers in the app.
GET STARTED
From download to your first words.
Choose a starting setup. Change providers whenever your needs change.
- 01
Download the app
Get Vowrite and follow the installation steps for your platform.
- 02
Set up your providers
Add your API keys on the app’s API Keys page, or set up a supported local model for refinement. Refinement is optional.
- 03
Speak, then insert
Use your voice shortcut, speak, and insert the text into the app you are using.
What determines cloud costs?
Your usage and the model’s rates determine the bill. A different supported model can change the cost. Check its quality, availability and billing terms before switching.
GOOD TO KNOW
Clear costs. Clear choices.
Who am I paying?
Vowrite is free and open source. Cloud API charges go directly to your chosen providers. Their rates, minimum charges and credit terms apply.
Can I change providers later?
Yes. Choose a supported provider in the app and configure its API key or local model. Recognition and refinement can use different providers.
How are recognition and refinement billed?
Cloud recognition is billed for audio; refinement is billed for input and output text tokens. Exact rules depend on the provider. You can skip refinement.
When does local processing avoid cloud fees?
A locally processed step makes no cloud API call and incurs no cloud API fee. Local models currently cover text refinement (Ollama or MLX Server on Mac); speech recognition still uses a cloud provider. Local models need compatible hardware and setup; your device and electricity are separate costs.
What if I only use it occasionally?
Vowrite has no monthly subscription to maintain. Cloud charges follow your actual billable usage, subject to your provider’s minimums and terms.
See current provider rates.
Current terms always come from the provider.
Full provider reference · July 2026Preserved tables and examples; not current offers
Historical reference, not current offers. Original estimates, model details and notes are preserved below. Promotions may have expired; verify current prices and availability with the provider.
Original provider tables are preserved below as an archival snapshot, not verified current quotes or current Vowrite capabilities. In particular, Sherpa is unavailable and Ollama / MLX provide text refinement only. Use the official billing links for current rates.
The original model and pricing notes below are preserved in English.
Provider Pricing
Vowrite is free and open source. You bring your own API keys (BYOK) and only pay your chosen providers directly. Here's what they cost.
Archived monthly examples · July 2026
Old assumptions, not a forecast for your usage. Prices, model availability and free-tier claims below have not been reverified for today.
Monthly Cost Estimates
Based on ~30 minutes of voice input per day (≈15 STT hours + 300K Polish tokens / month)
🆓 Free Combo
SiliconFlow SenseVoice STT + Zhipu glm-4.7-flash Polish. Zero cost, unlimited usage.
💰 Budget
Groq whisper-large-v3-turbo + DeepSeek v4-flash Polish. Excellent quality at pennies.
⚖️ Balanced
OpenAI mini-transcribe + gpt-5.4-mini. Reliable mainstream choice.
🎯 Premium
OpenAI gpt-4o-transcribe + Claude Sonnet 5. Top accuracy and polish quality.
🎤 Speech-to-Text (STT)
Transcribe your voice to raw text
| Provider | Model | Price / min | Price / hour | Languages | |
|---|---|---|---|---|---|
| Sherpa (Offline) | SenseVoice Small / Zipformer / FireRedASR2 | Free 🔒 | Free | EN/CN/JP/KR/方言 | 100% LOCAL |
| SiliconFlow | FunAudioLLM/SenseVoiceSmall | Free ✨ | Free | 50+ | FREE ⭐ |
| SiliconFlow | TeleAI/TeleSpeechASR | Free ✨ | Free | CN focused | FREE |
| iFlytek (讯飞) | iat / xfime-mianqie | Free 500/day | ~$0.002 | CN + 23 dialects | CN BEST |
| Groq | whisper-large-v3-turbo | $0.0007 | $0.04 | 57+ | FASTEST ⭐ |
| Groq | whisper-large-v3 | $0.0019 | $0.111 | 57+ | Higher accuracy |
| OpenRouter | openai/whisper-large-v3 | $0.0015 | $0.09 | 57+ | ⭐ Default |
| OpenRouter | qwen/qwen3-asr-flash-2026-02-10 | $0.0021 | $0.126 | 11 · strong Chinese | |
| Qwen | qwen3-asr-flash | ~$0.0028 | ~$0.17 | 30+ auto-detect | ⭐ Default |
| Qwen | qwen3-asr-flash-realtime | — | — | 30+ streaming | Pricing unverifiable 2026-07 |
| Qwen | paraformer-realtime-v2 | ~$0.0014 | ~$0.08 | 30+ | Legacy |
| Qwen | fun-asr | — | — | Multilingual | Pricing unverifiable 2026-07 |
| Together AI | whisper-large-v3 | $0.0015 | $0.09 | 57+ | ⭐ Default |
| Together AI | nvidia/parakeet-tdt-0.6b-v3 | $0.0015 | $0.09 | EN/EU only | No Chinese |
| OpenAI | gpt-4o-mini-transcribe | $0.003 | $0.18 | 57+ | ⭐ Default |
| OpenAI | gpt-4o-transcribe | $0.006 | $0.36 | 57+ | ACCURATE |
| OpenAI | gpt-4o-transcribe-diarize | $0.006 | $0.36 | 57+ | Speaker labels |
| OpenAI | whisper-1 | $0.006 | $0.36 | 57+ | Legacy |
| Deepgram | Nova-3 | $0.0077 | $0.46 | 36+ (zh mono only) | TOP ACCURACY ⭐ |
| Deepgram | Nova-3 Medical | — | — | Medical domain | See deepgram.com/pricing |
| Deepgram | Nova-2 | $0.0058 | $0.35 | 36+ | Legacy |
| Volcengine | Doubao 录音文件识别 | — | $0.118 | 13+ | Not yet integrated (TOS-bucket upload) |
✨ AI Polish (LLM)
Clean up and refine your transcribed text
| Provider | Model | Input $/M tokens | Output $/M tokens | Context | |
|---|---|---|---|---|---|
| Ollama (Local) | qwen3:8b · qwen3.5:9b · llama3.1:8b · deepseek-r1:8b | Free 🖥️ | Free | Varies | 100% LOCAL |
| MLX Server (Local) | Qwen3.5-9B · Llama 3.3 70B · Mistral Small 24B | Free 🖥️ | Free | Varies | APPLE SILICON |
| SiliconFlow | Qwen/Qwen3.5-4B | Free ✨ | Free | 256K | FREE |
| SiliconFlow | Qwen/Qwen3-8B | Free ✨ | Free | 32K | FREE |
| Zhipu | glm-4.7-flash | Free ✨ | Free | 203K | FREE ⭐ |
| Cerebras | gpt-oss-120b | Free ✨ | Free (rate-limited) | 128K | FREE ⭐ ~3000 tok/s |
| Groq | openai/gpt-oss-120b | $0.15 | $0.60 | 128K | ⭐ Default |
| Groq | openai/gpt-oss-20b | $0.075 | $0.30 | 128K | |
| Groq | llama-3.1-8b-instant | $0.05 | $0.08 | 128K | FASTEST |
| Groq | qwen/qwen3-32b | $0.29 | $0.59 | 128K | Reasoning · Preview |
| Groq | llama-3.3-70b-versatile | $0.59 | $0.79 | 128K | |
| DeepSeek | deepseek-v4-flash | $0.14 | $0.28 | 1M | VALUE ⭐ |
| DeepSeek | deepseek-v4-pro | $0.435 | $0.87 | 1M | |
| OpenAI | gpt-5.4-nano | $0.20 | $1.25 | 400K | CHEAPEST |
| OpenAI | gpt-5.4-mini | $0.75 | $4.50 | 400K | ⭐ Default |
| OpenAI | gpt-5.6-luna | $1.00 | $6.00 | 1M | New gen · fast |
| OpenAI | gpt-5.6-sol | $5.00 | $30.00 | — | FRONTIER |
| Gemini | gemini-2.5-flash-lite | $0.10 | $0.40 | 1M | CHEAPEST |
| Gemini | gemini-3.1-flash-lite | $0.25 | $1.50 | 1M | New gen |
| Gemini | gemini-2.5-flash | $0.30 | $2.50 | 1M | ⭐ Default |
| Gemini | gemini-3.5-flash | $1.50 | $9.00 | 1M | New gen · fast frontier |
| Gemini | gemini-2.5-pro | $1.25 | $10.00 | 1M (2× >200K) | FLAGSHIP |
| Claude | claude-haiku-4-5 | $1.00 | $5.00 | 200K | Fast tier |
| Claude | claude-sonnet-5 | $2.00 * | $10.00 * | 1M | PREMIUM ⭐ * intro price to 2026-08-31, then $3/$15 |
| Claude | claude-opus-4-8 | $5.00 | $25.00 | 200K | Most capable |
| Kimi (Moonshot) | kimi-k2.5 | $0.60 | $3.00 | 200K | Cheaper · final availability 2026-08-31 |
| Kimi (Moonshot) | kimi-k2.6 | $0.95 | $4.00 | 200K | ⭐ Default |
| MiniMax | MiniMax-M3 | $0.30 | $1.20 | 197K | ⭐ Default · only thinking-off model |
| MiniMax | MiniMax-M2.7 | $0.30 | $1.20 | 197K | Thinking always on |
| MiniMax | MiniMax-M2.7-highspeed | $0.60 | $2.40 | 197K | FAST thinking on |
| Volcengine | doubao-seed-1-8 | from $0.11 | from $0.28 | 32K | ⭐ Default |
| Volcengine | doubao-seed-1-6-flash | from $0.02 | from $0.21 | — | Cheapest, fastest |
| Volcengine | doubao-seed-2-1-turbo | $0.42 | $2.08 | — | New gen · balanced |
| Volcengine | doubao-seed-2-1-pro | $0.83 | $4.17 | — | FLAGSHIP |
| Qwen | qwen3.7-plus | from $0.40 | from $1.60 | 1M | ⭐ Default · tiered |
| Qwen | qwen3.7-max | $2.50 | $7.50 | — | FRONTIER replaces 3.6-max-preview |
| Qwen | qwen3.6-flash | from $0.25 | from $1.50 | — | VALUE |
| SiliconFlow | deepseek-ai/DeepSeek-V3 | ~$0.014 | ~$0.028 | 164K | CHEAPEST ⭐ |
| SiliconFlow | deepseek-ai/DeepSeek-V3.1-Terminus | ~$0.28 | ~$0.42 | 164K | |
| SiliconFlow | zai-org/GLM-4.6 | ~$0.42 | ~$0.84 | 200K | |
| SiliconFlow | Qwen/Qwen2.5-72B-Instruct | ~$0.57 | ~$0.57 | 128K | |
| Zhipu | glm-5.2 | — | — | 1M | Newest — see bigmodel.cn/pricing |
| Zhipu | glm-5.1 | $0.45 | $0.80 | 200K | FLAGSHIP |
| Zhipu | glm-4.6 | $0.60 | $2.20 | 200K | |
| xAI (Grok) | grok-4.3 | $1.25 | $2.50 | 1M | ⭐ Default |
| xAI (Grok) | grok-4.5 | $2.00 | $6.00 | <200K | Flagship · thinking on |
| Cerebras | zai-glm-4.7 | — | — | — | Preview — see cerebras.ai/pricing |
| Cerebras | gemma-4-31b | — | — | — | Preview |
| Baidu Qianfan | ernie-4.5-turbo-128k | $0.11 | $0.44 | 128K | ⭐ Default · ¥0.8/¥3.2 |
| Together AI | LiquidAI/LFM2.5-8B-A1B | $0.03 | $0.12 | — | CHEAPEST PAID |
| Together AI | Qwen/Qwen3.5-9B | $0.17 | $0.25 | 32K | Cheap & fast |
| Together AI | Qwen/Qwen3.7-Plus | $0.32 | $1.28 | 1M | New Qwen gen |
| Together AI | meta-llama/Llama-3.3-70B-Instruct-Turbo | $1.04 | $1.04 | 128K | ⭐ Default |
| Together AI | deepseek-ai/DeepSeek-V4-Pro | $1.74 | $3.48 | 512K | Reasoning flagship |
| OpenRouter | 400+ models (aggregator) | Pass-through | Pass-through | Varies | One key, many models |
📌 Notes: CNY prices converted at ~¥7.2 = $1 USD (July 2026). Prices sourced from official provider pages and may change — verify on the provider's billing page before high-volume usage. Monthly estimates assume ~15 hours STT + ~300K tokens Polish per month. Models marked ⭐ are the Vowrite default for that provider. Cells marked "—" indicate pricing not yet published or unverifiable at time of writing — check the provider's official page.
Tables scroll horizontally to keep every column available.
Original reference links
Last updated: July 2026 · v0.2.3.0
Start with your voice.
Download Vowrite. Choose your providers. Make it yours.
Download Vowrite Free app · Cloud API usage billed separately