Scam With Your Friends supports voice cloning through WAV files. To work on an existing caller, use Daily Callers; to begin a new caller, use Character Creator. Hotfix 2 changed these entry points and removed the separate Localize Voice buttons.
A complete current import sequence, accepted sample rate, channel count, audio duration, and caller-file reference format have not been verified here. WAV support alone is not enough to invent those settings or promise that any WAV file will work.
Use the current caller-editing route
Decide first whether the voice belongs to a caller that already exists. Existing callers are edited through Daily Callers. Character Creator now starts directly in caller creation, so it should not be treated as a guaranteed shortcut to editing a saved character.
Open the relevant current interface and read the controls it actually provides. An older tutorial may ask you to press a separate Localize Voice button. Those buttons have been removed, so their absence is consistent with the updated interface rather than proof that your installation is broken.
The update does not supply every replacement control or a complete file-import path. Follow the current visible interface instead of substituting an invented button sequence.
Earlier caller-selection interface shown for orientation; this is not a WAV import screen, and current controls may differ.
Keep the supported format separate from unverified settings
| Item | What is confirmed |
|---|---|
| Custom voice source category | WAV files are supported for voice cloning |
| Existing caller editing | Daily Callers is the current entry point |
| New caller creation | Character Creator opens directly into creation |
| Separate Localize Voice buttons | Removed in Hotfix 2 |
| Required WAV sample rate, channels, or duration | Not verified here |
| Save directory and complete import sequence | Not verified here |
A WAV file can contain audio with different technical properties. Renaming another file's extension to .wav does not establish that its contents are a valid WAV file, and a valid WAV container does not prove that the game accepts every encoding or recording configuration.
Use the requirements displayed by the actual game when they are available. If the interface reports that a file is unsupported, preserve the message before changing settings. No specific conversion preset is established as a universal fix.
Preserve the caller before experimenting
Keep an unchanged copy of local material you intend to retain before editing it. Make one change at a time so you can distinguish a file problem from a change in the caller's configuration.
This is general file-management advice. It is not a claim that the game stores voices in a particular folder or automatically backs up every custom caller. Do not place the audio in a guessed directory based on another game's modding instructions.
Likewise, do not add a guessed voice field to a personality JSON file. The game supports JSON personalities and WAV voice cloning, but their exact file relationship has not been verified. Use an actual current-game file or the interface's own controls as a reference when available.
Separate custom voices from Qwen3 readiness
A custom caller's voice file and the game's Qwen3 voice preparation are different parts of the audio workflow. An unfinished Qwen3 download does not, by itself, tell you whether a custom WAV was imported correctly.
Launch guidance recommends that non-English players let Qwen3 text-to-speech finish downloading on the main menu for higher-quality speech. The game can also warn before hosting or joining when higher-quality voices are not ready, offering options to wait or continue.
Hotfix 2 fixed stalled caller voice loading and Qwen3 startup blocking caller input. It reduced memory use when loading custom caller voices and added clearer Qwen3 troubleshooting messages, including a reminder that Qwen3 can be disabled if too slow.
Those are specific improvements. They do not establish a fixed download time, a guarantee that every voice loads immediately, or a required hardware upgrade whenever speech is delayed.
Check whether the caller remains connected to Daily Callers
An earlier fix addressed uploaded custom callers losing their Daily Callers connection after a restart. If the caller disappears from that selection, distinguish that symptom from a voice file that is present but silent.
Workshop upload behavior is separate again. A fix addressed stuck Workshop uploads that blocked other caller uploads. Re-encoding audio is not an established solution to every stalled upload, and rewriting a personality is not proof that a missing caller-selection connection will return.
Identify the exact failure before changing the audio
Record which stage failed: selecting a file, opening a caller, loading the voice, producing spoken playback, maintaining the Daily Callers selection, or uploading to Workshop. Include the exact visible error when there is one.
If the AI displays a temporary usage restriction or failure status, treat that as system information rather than immediately blaming the WAV file. Current updates make those states explicit. A successful custom-voice import does not bypass AI-service restrictions, and a service interruption does not prove that the audio asset is corrupt.
The confirmed route is to work through the current caller interface with a WAV voice asset. More specific import instructions require the actual current controls and format requirements; this guide does not replace those missing details with a fictional recipe.