Save Drama 3 Voice:
Save Drama 3 Voice preserves an encoded voice reference so you can reuse it in later Drama 3 workflows without encoding the same reference again. It saves the voice’s audio codes—the numerical representation of the reference speech—together with its transcript in a single .safetensors file under output/drama3_voices. The codes are stored as contiguous 16-bit integers, while the transcript is stored as file metadata, keeping both parts together for Load Drama 3 Voice to read. This is useful for building an organized collection of reusable voice references for recurring characters or projects. The node saves an already encoded reference; it does not record speech, create a transcript, generate new dialogue, or export a playable audio file.
Save Drama 3 Voice Input Parameters:
voice
A required DRAMA3_VOICE connection containing the encoded voice reference you want to save. The supplied voice must contain both codes, a tensor holding the encoded audio information, and transcript, the corresponding text. Connect a compatible voice source, such as Drama 3 Encode Voice or Load Drama 3 Voice; a raw audio connection cannot replace this encoded voice connection. During saving, the node converts the codes to contiguous 16-bit integers and stores the transcript as metadata without editing its text. The content supplied here determines which reference is preserved, so check that the transcript matches the reference speech before saving. No default value, numeric minimum, numeric maximum, or selectable options are defined for this parameter.
name
A required STRING that sets the saved file’s name prefix and can also specify a subfolder within the voice output location. The default is voice; no minimum length, maximum length, or fixed list of options is specified in the node schema. The node adds a five-digit counter, a trailing underscore, and the .safetensors extension automatically, producing a filename such as voice_00001_.safetensors. For example, characters/narrator places the file in a characters subfolder under output/drama3_voices, with a name such as narrator_00001_.safetensors. The counter is selected by ComfyUI’s save-path handling, so its actual value depends on the destination’s existing files. This parameter affects file organization and naming, not the encoded reference or transcript.
Save Drama 3 Voice Output Parameters:
There are no output parameters or connection sockets. This is an output node: its result is a saved .safetensors file on disk rather than a value passed to another node. The file contains the encoded voice codes and transcript and can be reopened with Load Drama 3 Voice. No file path, audio preview, or success value is returned through a node output.
Save Drama 3 Voice Usage Tips:
- Encode your reference first, then connect the resulting
DRAMA3_VOICEto this node. For later sessions, use Load Drama 3 Voice to reuse the saved reference instead of repeating the encoding step. - Use descriptive prefixes such as
narrator_calmorcharacters/alex. Subfolders help separate projects or characters, while the automatically added counter distinguishes repeated saves. - Enter a prefix rather than a complete filename. For example, use
narratorinstead ofnarrator.safetensors, because the node appends its own counter and extension. - Check the reference and its matching transcript before saving. This node preserves the supplied content; it does not transcribe audio, correct text, or improve the voice reference.
- Save references you intend to reuse rather than repeatedly saving the same voice during experimentation. The node writes a file whenever it executes, and repeated executions can create additional numbered files.
- Use a separate audio-saving node if you need a playable recording. A saved Drama 3 voice file is an encoded reference for compatible Drama 3 nodes, not a WAV or MP3 export.
Save Drama 3 Voice Common Errors and Solutions:
KeyError: 'codes'
- Explanation: The supplied voice data does not contain the encoded audio codes that this node needs to save. This can occur if another custom node supplies an incomplete or incorrectly structured voice object.
- Solution: Reconnect a valid
DRAMA3_VOICEfrom Drama 3 Encode Voice or Load Drama 3 Voice. If you are creating the voice reference from audio, complete the encoding step before saving.
KeyError: 'transcript'
- Explanation: The supplied voice data does not contain a transcript. The save operation requires this text because it is stored alongside the codes for later loading.
- Solution: Check the upstream voice-creation workflow and recreate the reference with its transcript included. Supply a complete encoded voice rather than audio codes alone.
AttributeError: '…' object has no attribute 'to'
- Explanation: This possible Python error means the supplied
codesvalue does not support the tensor conversion used by the node. For example, a list or text value cannot be saved through this implementation in place of the expected PyTorch tensor. - Solution: Use the encoded voice produced by a compatible Drama 3 node without manually replacing its codes. If a custom conversion node is involved, bypass it and try saving the original voice connection.
Transcript metadata type error
- Explanation: The saving library expects metadata values to be strings. If an upstream node supplies a transcript as a list, number, or another non-text value, saving may fail with a type-conversion error; the exact message depends on the library version.
- Solution: Ensure that the upstream voice contains its transcript as plain text. Prefer the standard Drama 3 voice-encoding workflow, and correct any custom processing that changes the transcript into a different data type.
Permission denied or output folder is not writable
- Explanation: A filesystem error, such as
PermissionErroror an error containingPermission denied, can occur when ComfyUI cannot create the destination folder or write the voice file. The exact wording depends on your operating system and the saving library. - Solution: Check that the account running ComfyUI can write to
output/drama3_voicesand any requested subfolder. If you use a container or shared drive, check its write permissions and mount settings. Try the defaultnamevalue,voice, to rule out a problematic subfolder.
No space left on device
- Explanation: A disk-writing error containing wording such as
No space left on devicecan occur when the destination storage is full. The file may not have been saved completely. - Solution: Free space on the drive containing ComfyUI’s output folder, then run the node again. If the failed attempt left an incomplete file, remove it before trying to load that reference.
Invalid filename or path error
- Explanation: An error such as
Invalid argumentor another path-related failure may occur ifnamecontains characters or a path that your operating system does not allow. Subfolders are supported, but their names must still be valid on your system. - Solution: Try a simple prefix such as
voice_test. Use letters, numbers, underscores, and ordinary subfolder names, and avoid operating-system-reserved characters or names. Let the node add the counter and file extension automatically.
