AI Post-processing

AI post-processing uses a Large Language Model (LLM) to refine ASR transcripts. Typical improvements include removing filler words, fixing typos, adjusting punctuation, and polishing tone, making voice typing smoother and more natural.

Overview

Pipeline

Recording → ASR → [AI post-processing] → Insert text

Recommended use cases

Good for

Speech to writing: meeting notes, reports
Long-form input: reduce manual edits afterwards
Professional content: more formal/consistent output
Multilingual: combine with translation prompts for cross-language voice input

Not recommended

Casual chat (spoken style may feel more natural)
Very short input (single word, numbers)
Latency-sensitive scenarios

Supported LLM Providers

BiBi Keyboard supports 11 LLM providers. All of them use an OpenAI-compatible API format:

Vendor	Default model	Notes	Sign-up link
SF_FREE (SiliconFlow Free)	Qwen/Qwen3-8B	🆓 free, no config	https://cloud.siliconflow.cn/i/g8thUcWa
DEEPSEEK	deepseek-chat	💰 good value; reasoning mode	https://platform.deepseek.com/
ZHIPU	glm-4.6	🇨🇳 China-based; reasoning mode	https://bigmodel.cn/usercenter/proj-mgmt/apikeys
MOONSHOT	kimi-k2-0905-preview	🧠 long context; reasoning mode	https://platform.moonshot.cn/console/api-keys
VOLCENGINE	doubao-seed-1-6-flash	🇨🇳 Doubao; reasoning mode	https://console.volcengine.com/ark
OPENAI	gpt-4o-mini	🌍 OpenAI	https://platform.openai.com/signup
GEMINI	gemini-2.0-flash	🚀 fast; reasoning supported	https://aistudio.google.com/apikey
GROQ	llama-3.3-70b-versatile	⚡ very fast inference	https://console.groq.com/keys
CEREBRAS	llama-3.3-70b	⚡ fast inference	https://cloud.cerebras.ai/platform
OHMYGPT	gpt-4o-mini	🔀 multi-vendor relay	https://x.dogenet.win/i/CXuHm49s
CUSTOM	user-defined	🛠️ any OpenAI-compatible API	-

Reasoning mode

Some providers expose a "thinking/reasoning" mode. The model reasons before producing output, which can help for complex editing but usually increases latency and token usage.

Prompt Presets

BiBi Keyboard includes 5 built-in prompt presets and supports custom ones.

Built-in presets

Preset name	Use case	Effect
General post-process	daily voice input	remove filler words, fix slips, keep original meaning
Basic polishing	formal rewrite	grammar fixes, punctuation, smoother expression
Translate to English	cross-language	translate transcript into English
Extract key points	meeting notes	extract key info into a bullet list
Extract to-dos	task tracking	identify tasks and generate a checklist

Custom prompts

Go to Settings → AI Post-processing → Prompt presets:

Tap "Add preset"
Write your prompt (role, task, rules, output format, etc.)
Save and apply quickly in AI Edit

Configuration

Basic

Key	Type	Default	Description
`postProcessEnabled`	Boolean	`false`	master switch
`llmVendor`	LlmVendor	`SF_FREE`	selected LLM vendor
`llmEndpoint`	String	vendor default	API endpoint (auto for built-in vendors)
`llmApiKey`	String	`""`	API key (not needed for free service)
`llmModel`	String	vendor default	model name
`llmTemperature`	Float	`0.2`	temperature (0-2; lower = more deterministic)

Advanced

Key	Type	Default	Description
`postprocSkipUnderChars`	Int	`0`	skip AI post-processing if shorter than this (0=disable)
`activePromptId`	String	`""`	active prompt preset id
`promptPresetsJson`	String	`""`	prompt preset list JSON

Temperature hints

0 - 0.3: highly consistent, good for precise edits
0.4 - 0.7: balanced creativity and stability
0.8 - 2.0: more creative but less stable

Quick Setup

Use free service (recommended)

BiBi Keyboard ships with SiliconFlow free service. No API key is required:

Open Settings → AI Post-processing
Enable "AI post-processing"
Ensure vendor is SF_FREE (default)
Pick a prompt preset (recommended: "General post-process")
Tap the magic wand button on the keyboard to enable AI post-processing mode
Done — transcripts will be refined automatically

About the free service

Has a free quota (see SiliconFlow website for details)
Example models: Qwen/Qwen3-8B, THUDM/GLM-4-9B, etc.
If you need other models, register on SiliconFlow and use your own API key

Configure a paid vendor

Example with DeepSeek:

Sign up at https://platform.deepseek.com/
Create an API key and add credits
In Settings → AI Post-processing:
- Vendor: DEEPSEEK
- API key: paste your key
- Model: e.g. deepseek-chat
- Temperature: recommended 0.2
Save and test

Configure a custom vendor

For any OpenAI-compatible API:

Vendor: CUSTOM
Fill in:
- Endpoint: e.g. https://your-api.com/v1
- API key
- Model: e.g. gpt-3.5-turbo
- Temperature: recommended 0.2
Save and test

Custom endpoint requirements

Must be compatible with OpenAI Chat Completions API
The path is typically /v1/chat/completions (the app will append it automatically)

Reasoning Mode Support

Different vendors control reasoning mode in different ways:

Vendor	Control method	Supported models	Notes
DEEPSEEK	model choice	deepseek-reasoner	choose the reasoner model
MOONSHOT	model choice	kimi-k2-thinking	choose the thinking model
SF_FREE	toggle param	Qwen3 series, DeepSeek-V3.1, etc.	enable "Reasoning mode" in settings
GEMINI	toggle param	gemini-2.5-flash+	`reasoning_effort` param
GROQ	toggle param	qwen3-32b, gpt-oss series	`reasoning_effort` param
CEREBRAS	toggle param	gpt-oss-120b	`reasoning_effort` param
VOLCENGINE	toggle param	doubao-seed series, deepseek	`thinking.type` param
ZHIPU	toggle param	glm-4.6, glm-4.5 series	`thinking.type` param
OHMYGPT	toggle param	gemini-2.5, claude, gpt-5 series	`reasoning_effort` param

When to use reasoning mode

✅ complex rewrites (technical terms, strict formatting)
✅ tasks needing reasoning (e.g. to-do extraction)
✅ multi-step transformations (e.g. translate + polish)
❌ simple filler-word removal (extra latency without benefit)

Tips

There are three ways to trigger AI post-processing:

Trigger	Description	Best for
Auto post-process	runs automatically after each voice input	daily use; output is final
AI Edit	select text and open AI Edit; choose a prompt preset for the edit	iterative edits / retrying
Skip short input	set a minimum length threshold to skip AI for short phrases	avoid overhead on tiny inputs

Troubleshooting