On Maloum (app.maloum.com), a voice message does something a wall of text can't: it sounds like a person. Fans reward that. But for an agency, voice has always been the hard part to scale, because it seems to require the model to sit down and record on demand, in the fan's language, at the exact moment a conversation is heating up. That doesn't happen at scale, so most agencies simply skip voice and leave conversion on the table.
PinkForge changes the constraint. It's a Chrome Extension and dashboard for creator agencies, and its voice cloning feature lets your chatters send voice messages in the model's own voice, in any language, without the model recording anything in the moment. This guide walks through exactly how to do it inside the Maloum chat, and where it fits in a scaled operation.
The one-line version: record a 60-second voice sample once, and from then on your team can generate voice messages in that model's voice, in any of 15+ languages, directly in Maloum.
Text can be translated, batched, and sent by anyone on your team. Voice feels personal, immediate, and hard to fake, which is exactly why it lands. The friction has always been production: a model can't be on call to record a custom note for every fan in every language, and pre-recorded generic clips get stale fast and can't respond to what a specific fan just said. So voice becomes a "nice to have" that never gets used consistently. Voice cloning removes that friction by separating the model's voice from the model's availability.
For an agency, the operational consequence is what matters. A model's time is your scarcest resource — it goes to content, not to sitting in a booth recording bespoke audio for individual fans. Every voice note that requires the model in the loop is a note that mostly won't get sent. The moment you decouple the voice from the person's schedule, voice stops being a special-occasion gesture and becomes a standard tool your chatters reach for whenever a conversation calls for a warmer, more personal beat. That shift — from rare to routine — is the whole reason voice cloning is interesting to an agency rather than to an individual creator.
It also matters across languages. A fan in Germany, a fan in France and a fan in Brazil all respond better to a voice note in their own language than to one in English they half-understand. Recording that manually would mean the model speaking three languages fluently, which almost never happens. Generating it means the same model voice, three languages, zero extra recording sessions.
PinkForge installs as a Chrome Extension, so there is nothing to host. Once installed and signed in, it activates directly inside app.maloum.com. Setup takes about three minutes, and the 7-day free trial needs no credit card. Because the integration is native to Maloum, your chatters work in the same inbox they already use.
This is a one-time setup per model. The model records a single 60-second voice sample, which PinkForge uses to build the cloned voice. You do not need a studio; a clean phone recording in a quiet room is enough to capture the voice. After this step, the model never has to record on demand again.
Consent matters. Only clone a voice you are authorized to use. The 60-second sample should come from the model themselves, with their agreement to use voice messages generated this way.
In the Maloum chat, write what the voice message should say. This is where the rest of PinkForge helps: real-time translation in 15+ languages means your chatter can type in their own language and have the voice message produced in the fan's language, and AI reply suggestions give them a fast starting point for what to say. The voice message is generated in the model's cloned voice regardless of the language, so an English-speaking chatter can send a German or Italian voice note that sounds like the model.
Generate the voice message and attach it to the conversation directly in the Maloum chat. There is no recording booth, no waiting on the model, and no switching to another app. The chatter stays in the same inbox they use for everything else. This is what makes voice usable at scale: any chatter, any language, any time, in the model's voice.
Voice notes should sound like the persona, not like a generic script. PinkForge gives each model its own glossary, so pet names, recurring phrases and persona-specific terms come out consistently. When several chatters can send voice in the same model's voice, the glossary is what keeps the persona coherent across the whole team.
Illustrative example (not real data): Say a model on Maloum has fans across Germany, France and Brazil, and the model can realistically record maybe a handful of custom voice notes per week. Without cloning, the other markets get none. With PinkForge, the model records one 60-second sample, and from then on chatters send voice messages in German, French and Portuguese in the model's voice whenever a conversation calls for it. If you want to run voice on its own before adding full translation, the Voice Only plan is 29€/month; full chatting plans start at 69€/month per model. See the AI voice messages explainer for how the cloning works in more depth.
Honestly, voice is not a magic button. If a fan is early in a conversation and hasn't warmed up, or if your persona is one where a voice note would feel off, text is still the better call. Voice cloning gives you the option to send voice at scale; it does not decide for you when to use it. Treat it as one strong tool in the chatter's kit, used when the moment fits, rather than something to spray at every fan. Used with judgment, it's one of the clearest ways to make a scaled operation still feel personal.
PinkForge runs as a Chrome Extension inside the Maloum chat. After the model records a one-time 60-second voice sample, your chatters can generate voice messages in that model's voice, in any of 15+ languages, and send them directly in the conversation.
No. The model records a single 60-second sample once. After that, chatters generate voice messages in the model's cloned voice on demand, so the model never has to record for individual conversations.
Yes. Combined with real-time translation across 15+ languages, a chatter can produce a voice message in the fan's language while it still sounds like the model, even if the chatter does not speak that language.
There is a Voice Only plan at 29€/month if you just want voice messages. Voice is also included in the full chatting plans, which start at Starter 69€/month per model. Every plan has a 7-day free trial with no credit card required.
Only clone a voice you are authorized to use. The 60-second sample should come from the model with their consent to send voice messages generated from it. Used with the model's agreement, it is a way to make scaled chatting feel personal.
Set up in 3 minutes. 7-day free trial. No credit card required.
Start Free TrialVoice Only - 29 EUR/mo