Which tool for which job
| You want to… | Use |
|---|---|
| Count tokens or estimate cost before sending a prompt | AI Token Counter |
| See whether a document fits a model’s context window | Context Window Calculator |
| Repair JSON that won’t parse | Broken JSON Fixer |
| Move a table from a chat into Excel or Google Sheets | Markdown Table to Excel |
| Remove zero-width or non-breaking spaces from copied text | Invisible Character Remover |
| Turn a YouTube or meeting transcript into readable paragraphs | Transcript Cleaner |
| Check which job requirements your resume proves | Resume vs Job Match |
| Get pixel sizes for an AI image or social post | AI Image Aspect Ratio |
| Paste a chat reply into Word or Google Docs without ** and # | ChatGPT to Word |
| Encode or decode Base64, JWT parts or URLs | Base64 & URL Encoder |
| Create a strong password privately | Password Generator |
Handy numbers
- 1,000 tokens is about 750 words of ordinary English. Code, numbers and most other languages use more tokens per word.
- Common hidden characters: zero-width space (U+200B), non-breaking space (U+00A0), narrow no-break space (U+202F), soft hyphen (U+00AD) and the byte order mark (U+FEFF).
- JSON allows only double quotes, no trailing commas and no comments, which is why text copied from a chat often fails to parse.
Private by design
Every tool in this section runs in your browser. Text, documents and data you paste are never uploaded, stored or used for training, which matters when you are working with resumes, client data or unpublished writing.
Common questions
Which model has the biggest context window? As of October 2026: Gemini 2.0 Pro at 2M tokens; Claude 4.x Enterprise at 1M tokens; GPT-5 variants up to 400K; standard Claude 4.x and GPT-5 at 200K and 128K respectively. Context works best in the first and last 25% of the window — the "lost in the middle" effect degrades retrieval across long contexts.
Why does AI output broken JSON? Models learn from code using single quotes, trailing commas and unquoted keys (JS/Python). Even with explicit "output valid JSON" instructions, drift happens. The JSON Fixer handles the common cases — trailing commas, single quotes, unquoted keys, unescaped newlines.
Are my inputs private? On every tool in this section, yes — processing runs in your browser as JavaScript. Nothing is uploaded, logged or sent to a server. Safe for resumes, client data and personal information. The site has no login, no account, no analytics on the text you paste.
What's a safe token budget per API call? Below 50K tokens of input is usually fine on any model. 50K–200K is OK but response quality can drift. 200K+ needs chunking or retrieval augmentation (RAG) to maintain accuracy. For code review or document QA, chunk by section rather than feeding monoliths.
How do I know if a resume is ATS-friendly? Avoid tables, columns, images, uncommon fonts, headers/footers with important content, and complex formatting. Use standard section headings ("Experience," "Skills," "Education"). Save as .docx or PDF (not .pages). Test with the Resume Job Match Checker — it flags common ATS issues.
What are tokens in simple terms? Chunks of text the model works with. English averages ~0.75 words per token, so 1,000 tokens ≈ 750 words. "ChatGPT" is 1 token; "unbelievable" might be 2–3 tokens; a Chinese character is often 1 token. Different providers use slightly different tokenizers.
Are AI model costs predictable? Yes, roughly. You pay per token in + token out at the provider's published rate. 10,000 input + 2,000 output tokens on Claude 4.x Sonnet = (10,000 × $3 + 2,000 × $15) ÷ 1,000,000 = $0.06. Caching the system prompt can cut costs significantly for repeat prompts.