Load Drama 3 Voice:
Load Drama 3 Voice restores a previously saved voice reference so you can reuse it in a Drama 3 speech-generation workflow without preparing and encoding the same reference recording again. It reads the saved reference’s transcript and encoded audio information, checks that both required components are present, and prepares the encoded information for use by other Drama 3 nodes. This makes it useful for keeping a reusable collection of character or narrator references across projects. It is designed to load files created by Save Drama 3 Voice in output/drama3_voices; it does not load ordinary audio recordings, generate speech, or load the Drama 3 model itself.
Load Drama 3 Voice Input Parameters:
voice
A required dropdown selection identifying the saved Drama 3 voice reference to load. Its options come from the files listed in ComfyUI’s registered drama3_voices folder, described by this extension as output/drama3_voices. Select a reference previously created with Save Drama 3 Voice; that node saves a .safetensors file containing encoded audio codes and transcript metadata. Your selection determines which reference is made available to downstream speech-generation nodes, but this loader does not change the reference’s voice characteristics or transcript. The selected file must contain both the transcript metadata entry and the codes tensor, or loading is rejected. Options are populated from the folder’s current file list rather than a fixed set of names. No explicit default is defined in the node schema, and minimum and maximum values do not apply.
Load Drama 3 Voice Output Parameters:
DRAMA3_VOICE
A custom DRAMA3_VOICE reference containing the saved transcript and encoded audio codes. The transcript is the text associated with the reference recording, while the codes are a numerical representation of that recording used by the Drama 3 workflow. Internally, these are supplied as transcript and codes, with the codes converted to PyTorch’s long integer type for downstream use. Connect this output to a compatible Drama 3 voice input, such as a voice slot on Drama 3 Generate. This is a reusable voice reference, not playable audio, newly generated speech, or a model; a speech-generation workflow is still needed to produce new audio.
Load Drama 3 Voice Usage Tips:
- Create reusable references with Save Drama 3 Voice first, then select the resulting file here. Renaming a normal audio file to
.safetensorswill not create a compatible voice reference. - Use descriptive names when saving references, such as a character name and recording variation. Save Drama 3 Voice adds a counter and supports subfolders, which can help you organize multiple references.
- Connect the loaded reference to the appropriate voice slot on Drama 3 Generate. That node associates slot
voiceNwith speaker N, so place each character’s reference in the slot corresponding to that character’s speaker tag. - If a newly saved reference is missing from the dropdown, check that it was saved under
output/drama3_voicesand refresh the available file list or reload the ComfyUI interface as needed. - Reuse saved references when you want to avoid repeating reference preparation and encoding. This node has no quality, sampling, or performance controls beyond selecting the file; speech-generation settings belong to downstream nodes.
- Prepare clean reference audio and an accurate matching transcript before saving the voice. Loading restores the saved information as it is; it does not trim audio, correct transcription mistakes, or improve a poor reference.
Load Drama 3 Voice Common Errors and Solutions:
{voice} is not a Drama 3 voice file
- Explanation: This is the node’s explicit validation error, with
{voice}replaced by the selected filename. It occurs when the loaded file has no metadata, lacks thetranscriptmetadata entry, or lacks thecodestensor. An ordinary audio file, model file, or unrelated tensor file is not a compatible voice reference. - Solution: Select a file created by Save Drama 3 Voice. If the reference is incomplete or incompatible, recreate the encoded reference and save it again with that node.
File not found or selected voice path cannot be resolved
- Explanation: The node must resolve the selected filename inside the registered
drama3_voicesfolder before loading it. Loading can fail if the file was deleted, renamed, or moved after selection. The exact error wording comes from ComfyUI’s file-path handling rather than this node. - Solution: Confirm that the file still exists in
output/drama3_voicesor the configured location for that folder, refresh the available list, and select an existing reference. If you moved the workflow to another installation, copy its saved voice files there as well.
File-loading or invalid file-format error
- Explanation: The file-loading utility may fail before the node can check the reference contents if the selected file is damaged, incomplete, or not in a supported format. The exact error message depends on the loader and the file.
- Solution: Use the original file saved by Save Drama 3 Voice rather than a renamed or manually edited file. If it was copied or downloaded, copy it again; if loading still fails, recreate and save the reference.
Permission denied or file cannot be opened
- Explanation: Loading requires ComfyUI to be able to read the selected file and access its containing folders. Operating-system permissions or storage-access restrictions can prevent this, even when the filename appears in the dropdown.
- Solution: Make sure the account running ComfyUI has read access to the file and its folders. If the reference is on an unavailable or restricted drive, restore access or place a readable copy in the configured voice folder.
Error while converting codes to an integer tensor
- Explanation: After checking that
codesexists, the node converts it using.long(). A manually constructed or incompatible file could contain an entry with that name that is not suitable for conversion. The node checks for the required entries but does not fully validate their structure. - Solution: Recreate the reference with the extension’s encoding and saving workflow rather than manually assembling its contents. A successfully loaded file with unsuitable code structure can also cause failures later in generation, so use references produced by Save Drama 3 Voice.
