Average cost per minute for AI voice agents built on Retell
Adding the component prices and one billing rule that’s easy to miss. On Retell’s list prices (checked 25 October 2026), a minute is the voice engine ($0.055, speech-to-text included), plus the LLM (GPT-5.6 Luna is $0.0064, GPT-5.6 Terra $0.064), plus the voice ($0.015 for Retell platform voices, $0.040 for ElevenLabs Flash), plus telephony ($0.015 a minute on a Retell-provided Twilio US local number; on your own SIP trunk Retell charges none and your carrier bills you instead). The rule: once…
Restaurant Voice Agent — Production tuning + scalable POS playbook (paid project)
Jon, are you still placing work for the restaurant rollout and POS playbook, or has this been delivered? I am Jason Ji, a US-based individual contractor with 12 years of professional development experience and 5+ years in production Node.js/Python work. I can own an agreed backend integration and release-testing workstream alongside your Retell specialist. For a reusable POS adapter, I would make menu/modifier mappings explicit, give orders stable operation IDs, reconcile accepted orders after…
Restaurant Voice Agent — Production tuning + scalable POS playbook (paid project)
I run a live Retell agent in production (voice, tools, and a custom LLM) at https://ced-castillo.com. Most of my time on it has gone to the same problems you listed: replies that run long, filler, awkward handoffs, and calls that fail when the path is not the happy one. I can review your current flow and give you a concrete list of changes: Where the conversation should be shorter, and which lines to cut How to structure the prompt, nodes, and transitions so the agent stops wandering What to…
We’re currently looking for an experienced Retell AI expert to review our existing voice agent workflow and provide feedback
I run a live Retell agent in production (voice, tools, and a custom LLM) at https://ced-castillo.com. Most of my time on it has gone to the same problems you listed: replies that run long, filler, awkward handoffs, and calls that fail when the path is not the happy one. I can review your current flow and give you a concrete list of changes: Where the conversation should be shorter, and which lines to cut How to structure the prompt, nodes, and transitions so the agent stops wandering What to…
Add OpenAI GPT-Live-1 (full-duplex speech-to-speech) support
Hi Retell team and community,I’d love to see GPT-Live-1 added as a speech-to-speech model option in Retell.Why it matters for naturalnessChatGPT’s Advanced Voice / GPT-Live experience feels significantly more human than most current voice agents. The main reasons: Full-duplex architecture — it can listen and speak at the same time. This enables natural backchanneling (“mhmm”, “yeah”), smoother interruptions, and better handling of pauses while someone is thinking. Better turn-taking — far fewer…
Webhook with dynamic variable not getting value
Hey @sdelinois If the email address is the same for every outbound call, the easiest option is to set it as an agent-level default dynamic variable. In your agent settings, find the Default Dynamic Variables section and add: email → name@gmail.com Then you can use it in your webhook URL like: ?email={{email}} Retell will use the agent-level default value whenever that variable is not supplied separately for the call. Retell documents this under Configure agent-level default dynamic variables.…
Webhook with dynamic variable not getting value
mark1: retell_llm_dynamic_variables Thanks for your help as I’m new to retell. So if I can’t pass a value in the prompt like {{email}}= “name@gmail.com” Where else can I set a variable that can be used in the webhook. Also, please point me to the steps or the documentation for accomplishing that. Thanks in advance for your help.
Webhook with dynamic variable not getting value
Hey @sdelinois Thanks for clarifying. For an outbound call, defining something like {{email}} = "name@gmail.com" inside the agent prompt does not assign a runtime value to that dynamic variable. For the webhook URL to resolve {{email}}, the value needs to be provided as an actual dynamic variable — either: as an agent-level default dynamic variable, or in retell_llm_dynamic_variables when creating the outbound call. For example, the call should contain an email value inside…
Webhook with dynamic variable not getting value
This is for an outbound call. Call Ended webhook. I want to send the transcript to the email that’s defined in the variable which is set in the agent prompt {{email}} = “name@gmail.com” I’ve used these variables for everything else but this one is not working for the webhook.
Webhook with dynamic variable not getting value
Hey @sdelinois Retell supports dynamic variables in both agent-level and account-level webhook URLs. For example: ?customer={{customer_name}} However, customer_name must already have a value for that call. Retell’s documentation states that if a dynamic variable has not been assigned, the placeholder remains unresolved rather than receiving a value automatically. For outbound calls, please make sure customer_name is included in retell_llm_dynamic_variables when creating the call. For inbound…
We’re currently looking for an experienced Retell AI expert to review our existing voice agent workflow and provide feedback
Hi @kkean, I saw your post about refining an existing Retell voice agent, particularly around edge cases, transfers and overall reliability. I’m developing OutcomeGuard, an approach to testing whether AI-driven workflows behave correctly under both normal and failure conditions. Rather than focusing only on prompt quality, I would suggest establishing a reproducible baseline with scenarios such as failed transfers, interrupted calls, invalid tool responses and incomplete downstream actions. The…
AI voice Consultant for refinement
Hi @mgilbert, I came across your post about Retell AI integration with backend databases and the difficulty of handling callers who deviate from expected conversation flows. I’m working on OutcomeGuard, an early-stage verification project focused on confirming that AI-triggered actions produce the correct downstream results. Your read/write integration is an interesting fit: an agent should not confirm a database update unless the operation actually succeeded and the resulting state is correct.…
Webhook with dynamic variable not getting value
webhook-overview documentation says says Webhook URLs support dynamic variables. For example, setting webhook_url to webhook?customer={{customer_name}} resolves the {{customer_name}} placeholder per call before the event is delivered. This applies to both agent-level and account-level webhook URLs. But when I try to do this no value is passed through. customer={{customer_name}} If I pass the value directly like it works. customer=John I removed URLs from post because it would not let me add them
GPT Live Integration
I completely agree! Combining the ultra-low latency of the new GPT Realtime model with a backend LLM for tool calling and knowledge base lookups is going to be an absolute game-changer for all of us. The pricing structure is just the cherry on top.
Voice Agent 'Time-Zone Smart' Scheduling
The Idea: Currently, outbound agents often call leads at inappropriate hours if the agent is based in a different time zone than the lead. This feature would allow Retell to automatically detect the lead’s location based on their area code (or CRM data) and “gate” the calling window. You simply set the allowed calling hours (e.g., 9:00 AM – 6:00 PM) in the agent settings, and Retell’s engine automatically pauses any pending outbound calls to that specific number until their local time falls…
Best LLM, Voice model, and Version?
Hey @marcus There isn’t one universally “best” LLM and voice combination, because the most natural experience depends on the balance between intelligence, latency, voice quality, and your specific prompt. For your use case — a relatively straightforward auto-repair receptionist that collects caller details and books appointments — I’d recommend starting with GPT-5.6 Terra. Retell currently marks it as a Recommended model, and Retell’s benchmark shows a strong balance of conversational…
Best LLM, Voice model, and Version?
For a receptionist bot, GPT-4o is solid and way cheaper than 4.1. Cartesia is genuinely better than ElevenLabs for naturalness — use their latest model. Honestly though, the voice model matters way less than your prompt and call flow for this use case. Get the basics dialed in first (clear instructions, good error handling), then swap models if needed. You’ll get 80% of the way there with any of these.
Best LLM, Voice model, and Version?
hello! im wondering what’s the best LLM model (gpt 4.1 vs 5.6 terra for example) and Voice provider (elevenlabs vs cartesia for example) and voice provider version (elevenlabs v3 vs v4 for example) possible? I want the absolute best most human natural sounding agent. My use case is an AI receptionist for an auto repair shop, it simply collect caller details (name, car, service, date/time) and books appointments. Thank you!
Restaurant Voice Agent - Looking to have an expert review our Retell setup
Craig, is the multi-venue booking review still open? One check I would add is a reservation trace that connects the Retell tool event to the actual booking record, venue and time zone. A tool’s success result alone does not establish that the reservation was saved correctly. Duplicate events, cancellations and a lost response after a successful write deserve separate tests. I build workflow software and integration checks. This packet shows a synthetic stateful workflow, including an…
Agent doesn't speak after pressing digit until customer speaks first
Hey @supernova77 The transfer itself doesn’t appear to be failing. Rather, the LLM is not consistently triggering the agent_transfer tool call alongside press_digit on some calls. Here are a few recommendations to improve consistency: Update the agent prompt: Add an explicit instruction such as: “When you hear ‘press N to take this call,’ you MUST trigger both press_digit and agent_transfer in the same turn. Never trigger only one.” Enable strict mode: You can turn on the “Structured output”…
Agent doesn't speak after pressing digit until customer speaks first
Hey @supernova77 I’m checking this with the team and will get back to you as soon as I have an update. Thank You
Agent doesn't speak after pressing digit until customer speaks first
Hi Shah, This call worked as expected and you can see botht he press digit tool call and the agent transfer tool call: call_d6746881d4827cc58c69736a193 on this call i only see the press 1 being pressed: call_501d6b2c1735fb42e3515fcec39 this one also only press but doesnt transfer: call_6d6931b11344161c372c905a756 My customer is complaining and we really need to resolve this issue. Your help is very appreciated. in the mean time i will see how to implement and look into the other option you had…

Warm transfer to a PBX call group: Is an internal handoff between groups considered a successful transfer
Hi @mark1, thanks for the quick reply. Following up on transfer ring duration, we primarily want to understand what counts as “answered” for a warm transfer, and whether internal handoffs could be considered a successful transfer Thanks again
How to Write ELITE Prompts for AI Voice Agents (With Examples)
Really solid tutorial. The distinction between single and multi-prompt agents is something I see a lot of people struggling with when they first start building. Using Miro to map out the flow before writing the prompt is such an underrated step. Thanks for the breakdown!
Warm transfer to a PBX call group: Is an internal handoff between groups considered a successful transfer
Hey @hussain Regarding dynamic variables in custom SIP headers, yes — Retell supports using dynamic variables as header values during transfers. Header names must begin with X-, or use User-to-User or Diversion. A couple of important notes: Custom SIP headers are preserved when transferring directly to a SIP endpoint, but they may be stripped when transferring to a PSTN number. For cold transfers using SIP REFER, some providers may ignore custom headers, including Twilio. For the Transfer Ring…
Agent doesn't speak after pressing digit until customer speaks first
Hey @supernova77 Could you please share your Agent ID and the Call IDs of the most recent failed calls? This will help us investigate the issue further. Thank You
This AI Improves Your Sales Agent Every Day
This workflow is excellent using Google Gemini’s large context window to process and summarize call transcripts is a smart move. The automated PDF generation and Slack reporting is exactly what agencies need to scale their analysis. Super clean build!
Warm transfer to a PBX call group: Is an internal handoff between groups considered a successful transfer
Hi Retell team, Wanted to confirm how a warm transfer works. Suppose we set the Transfer Ring Duration to 60 seconds. Suppose the destination is a call group on a clients phone system ending in 1000. The setup: Call group 1000 rings for 20 seconds. If no one answers, the phone system forwards the call to a call group 2000. My question: When the call moves from 1000 to 2000 on the phone system’s side, does Retell treat that as a successful transfer (the call was answered or connected)? Or does…
Agent doesn't speak after pressing digit until customer speaks first
Hi Shah, For some reason the workaround is no longer working and the agent is staying silent even after the call was transfered to the next agent. Do you know if the Delay before pick up has been implemented? Thanks, Leo