Custom Instructions
Inject your own AI prompts to extract domain-specific insights from every transcription. Build QA scorecards, compliance checklists, and custom analysis workflows.
How It Works
Pass an instructions field with your custom prompt (max 2,000 characters). The AI will process your prompt against the full transcript and return structured results in ai_analysis.custom_instructions_result.
curl
curl -X POST https://api.voxparse.com/v1/transcribe \
-H "X-API-Key: YOUR_API_KEY" \
-F "[email protected]" \
-F "instructions=Score this call 1-10 on: greeting quality, problem resolution, upsell attempt. Return JSON with scores and reasoning."Example Use Cases
| Use Case | Example Prompt |
|---|---|
| QA Scorecard | "Rate agent performance 1-5 on empathy, resolution speed, and script adherence" |
| Compliance Check | "Did the agent read the required disclosure? List any compliance violations" |
| Lead Qualification | "Extract: budget mentioned, timeline, decision maker name, next steps agreed" |
| Medical Notes | "Extract symptoms, diagnosis, prescribed medications, and follow-up instructions" |
Security
Injection protection: Custom instructions are sanitized server-side to prevent prompt injection attacks. Instructions attempting to override system behavior are silently dropped.
Maximum length: 2,000 characters. Included in the base rate at no extra cost.