TTS Evaluation - Trelis Studio

Model Evaluation

Model to Evaluate

HuggingFace model ID of a fine-tuned Orpheus model

Text Prompts

One prompt per line. If empty, uses dataset or default test prompts.

Or Use Dataset (optional)

Dataset with a text column. Prompts field takes priority if filled.

Max Samples

Limited to 50 samples. Sign in for more.

Speaker Name

Must match the speaker name used during training

Temperature

Top P

Repetition Penalty

Max New Tokens

HuggingFace Token (optional)

Required for private models or pushing results. Get token

Push results to HuggingFace as dataset

Saves generated audio samples as a playable HF dataset

Output Organization (optional)

Output Dataset Name (optional)

Make dataset private

No job

No evaluation running. Submit a job to see progress here.

Time	Model	Samples	Output	Cost	Status
No evaluations yet